FRAMEWORKS AND INTERFACES FOR UNLOADING DEVICE-BASED PACKAGE PROCESSING

专利摘要:
frameworks and interfaces for offload device-based package processing. high-speed packet processing to and from a virtualization environment can be provided while utilizing hardware-based segmentation offload and other such functionality. a hardware vendor such as a network interface card (nic) manufacturer may allow the hardware to support proprietary or open stateless tunneling in conjunction with a protocol such as single root i/o virtualization (sr-iov ) in order to implement a virtualized overlay network. the hardware can use various rules, for example, which can be used by nic to perform certain actions, such as encapsulating egress packets and de-encapsulating packets.
公开号:BR112013024883B1
申请号:R112013024883-1
申请日:2012-03-29
公开日:2021-06-29
发明作者:Pradeep Vincent；Matthew D. Klein；Samuel J. Mickelvie
申请人:Samuel J. Mckelvie J. Mckelvie；Amazon Technologies, Inc.；
IPC主号:

专利说明:

FUNDAMENTALS
[0001] As an increasing number of applications and services are made available over networks, such as the internet, an increasing number of content, application, and/or service providers are turning to multi-tenant shared resource technologies. Cloud computing, for example, can provide consumers with access to electronic resources through services, such as services on the Web, where the hardware and/or software used to support those services is dynamically scalable to meet the needs of the services at any time. given away. A consumer will typically rent, lease, or otherwise pay for access to resources through the cloud, and thus does not need to purchase and maintain the necessary hardware and/or software.
[0002] Such access comes with risks for providers of these shared resources, however, as there will typically be multiple users accessing the resources at various times. In cases where users have a virtual address space, so that the consumer's network functions as a single virtual network without the restrictions or additional addresses of one or more additional physical networks, it may be desirable to provide for the processing and routing of packets belonging to to that virtual address space. When consumers have access to devices, however, the routing and processing performance on a device can potentially allow the user to modify the routing or other such processing of packets. Furthermore, such functionality cannot be easily moved to many existing hardware devices that are not exposed to the user, for reasons such as size restrictions, protocol limitations, etc. BRIEF DESCRIPTION OF THE FIGURES
[0003] Various modalities in accordance with the present disclosure will be described with reference to the figures, in which:
[0004] FIG. 1 illustrates an environment in which various modalities can be implemented;
[0005] FIG. 2 illustrates an environment for providing access to various resources that can be used in accordance with a modality;
[0006] FIG. 3 illustrates a configuration for accessing specific hardware resources that can be used in accordance with a modality;
[0007] FIG. 4 illustrates a packet encapsulation process that can be used in accordance with an embodiment;
[0008] FIG. 5 illustrates the configuration for processing packets that can be used in accordance with an embodiment;
[0009] FIG. 6 illustrates an exemplary package header that can be used in accordance with various embodiments;
[0010] FIG. 7 illustrates an example of a fifth process for processing packets that can be used in accordance with various embodiments;
[0011] FIG. 8 illustrates an example of a sixth process for processing packets that can be used in accordance with various embodiments; and
[0012] FIG. 9 illustrates an exemplary flow for processing packets that can be used in accordance with various embodiments. DETAILED DESCRIPTION
[0013] Systems and methods conforming to various embodiments of the present disclosure may overcome one or more of the above-mentioned deficiencies or others experienced in conventional approaches to managing resources in an electronic environment. Systems and methods conforming to various embodiments provide for processing packets between a first address space, such as a virtual or consumer address space, and a second address space, such as a cloud network provider or address space physicist. Features, such as segmentation and desegmentation of commodity device offload characteristics, such as multiple network offload devices, can be used to help reduce overhead in relation to network traffic, particularly as it pertains to a virtualized environment. Various approaches to providing segmentation and desegmentation of discharge characteristics are described, for example, in co-pending US Patent application serial number 12/556,453 entitled "Stateless Packet Segmentation and Processing". filed "September 9, 2009", and application serial number 12/885,258 entitled "Framework for Stateless Packet Tunneling", filed September 17, 2010, each of which is incorporated herein by reference.
[0014] Various embodiments allow an offload device to support proprietary or open stateless tunneling in conjunction with a protocol such as single root I/O virtualization (SR-IOV) in order to implement a virtualized overlay network. SR-IOV generally refers to a standard specification for interoperability that allows a device such as a Peripheral Component Interconnect (PCI) device to appear as multiple independent physical devices. SR-IOV takes advantage of physical functions (PFs) and virtual functions (VFs). Physical functions are generally fully featured functions, while virtual functions are generally lighter weight functions that may lack at least some configuration features. SR-IOV typically requires support in the BIOS as well as support in the operating system instance or hypervisor running in hardware.
[0015] In at least some embodiments, a download device (or a supplier or manufacturer of such a device) can provide specific functionality for packet processing. For example, an implementation based on Dom-0 (ie, domain zero, typically the first domain started by the Xen hypervisor at boot time) might use various rules that can be used by an offload device to perform certain actions, such as how to encapsulate egress packets and unencapsulate ingress packets. Egress packet source checking can be performed on each of the egress packets based on VM source, including source MAC address and source IP address checking. In some embodiments, the offload device can enforce specific VLAN (virtual local area network) tags or otherwise add VLAN tags. After egress packet source verification, packets can be matched against a list of existing rules. In case there is a match, a corresponding encapsulation action can be taken on the packet and the packet transmitted accordingly. If not, the package can be sent to Dom-0's control software for further processing.
[0016] For ingress packets, packets in certain modalities can be identified as being encapsulated using a special format based, for example, on a predefined IP protocol number and a predefined one byte value at a predefined offset from of the L2 header end. These values can each be set by Dom-0. All ingress packets that are not encapsulated can be delivered to Dom-0. For encapsulated ingress, any opaque bits (located only after the outer L3 header) can be identified using a predefined length of opaque bits. Each packet can further be classified as belonging to a particular virtual machine (VM) (eg, an SR-IOV vector) using a one-byte field in the opaque bits at a predefined offset.
[0017] Each SR-IOV function can be configured with a set of ingress rules. Each rule can primarily consist of opaque bits to be matched with opaque bits of encapsulated ingress packets, an external source IP address, an external destination IP address, and source and target MAC addresses. When an ingress encapsulated packet matches one of the ingress rules for a particular SR-IOV function (ie, the opaque bits match), the packet can be unencapsulated (ie, the opaque bits are removed), the TTL of the The internal IP header is decremented by a value specified in the rule, and the packet is delivered to the VM corresponding to the SR-IOV function. Ticket packages that do not match any of the rules can be delivered to Dom-0.
[0018] In at least some embodiments, the offload device will maintain a packet count and a byte count for each encapsulation and de-encapsulation rule that could be read or reset from Dom-0. Various modalities can also provide the ability to inject packets into an SR-IOV function from Dom-0. Certain modalities may provide a debug mode in which each packet is forced through the irrespective Dom-0 of the matching rules that are in effect. A maximum transmission unit (MTU) for SR-IOV functions can be set from Dom-0 in at least one mode, defaulting to 1500. If and when a guest tries to change the MTU size, the device will discharge can ensure that the proposed MTU does not exceed the maximum MTU set by Dom-0. In some embodiments, the offload device may also perform connection tracking, which can be used to provide a session state firewall implementation on the offload device.
[0019] In at least some embodiments, Dom-0 control software can be provided, which manages encapsulation and de-encapsulation rules for both egress and ingress packets. The Dom-0 control software can manage the Address Resolution Protocol (ARP) cache for the substrate network, for example, using the packet count statistic provided by the offload device, as well as ARP queries from substrate. Dom-0's control software can also determine which rules, if any, should be pushed to the unloading device and which rules should be managed by Dom-0 as overflow rules in the case that the unloading device does not support all the rules that are needed.
[0020] FIG. 1 illustrates an example of an environment 100 for implementing aspects in accordance with various modalities. As will be appreciated, although a web-based environment is used for explanatory purposes, different environments can be used, as appropriate, to implement various modalities. The environment 100 shown includes both a development or test (or side) portion and a production portion. An electronic client device 102 may include any suitable device operable to send and receive requests, messages, or information over an appropriate network 104 and transmit information back to a user of the device. Examples of such client devices include personal computers, cell phones, handheld messaging devices, laptop computers, set-top box converters, personal data assistants, electronic book readers, and the like. The network may include any suitable network, including an intranet, the internet, a cellular network, a local area network, or any other such network or combination thereof. Components used for such a system may depend at least in part on the type of network and/or environment selected. Protocols and components for communicating over such a network are well known and will not be discussed in detail here. Communication across the network can be enabled by wired or wireless connections, and their combinations. In this example, the network includes the internet, as the environment includes a web server 106 to receive requests and serve content in response to them, although for other networks an alternative device serving a similar purpose could be used as would be evident to one skilled in the art. .
[0021] The illustrative environment includes at least one application server 108 and a plurality of resources, servers, hosts, instances, routers, switches, data stores, and/or other such components defining what will be referred to herein as a plan of 110 data, although it should be understood that the capabilities of this plan are not limited to storing and providing access to the data. It should be understood that there may be multiple application servers, layers, or other elements, processes, or components, which can be chained together or otherwise configured, that can interact to perform tasks such as getting data from storage appropriate data. As used herein, the term "data storage" refers to any device or combination of devices capable of storing, accessing and retrieving data, which may include any combination and number of data servers, databases, data storage devices. data, and data storage media, in any standard, distributed or clustered environment. The application server can include any appropriate hardware and software to integrate with data storage as needed to run aspects of one or more applications to the client device, handling a majority of the data access and business logic for an application. The application server provides admission control services in cooperation with data storage, and is capable of generating content such as text, graphics, audio, and/or video to be transferred to the user, which can be served to the user through from the web server in the form of HTML, XML, or other appropriate structured language in this example. In some embodiments, web server 106, application server 108, and similar components can be considered to be part of the data plane. The handling of all requests and responses, as well as the delivery of content between the client device 102 and the application server 108, can be handled by the web server. It should be understood that the web and application servers are not required and are merely exemplary components, as structured code can run on any appropriate device or host machine as discussed elsewhere here.
[0022] The environment also includes a development and/or test side, which includes a user device 118 allowing a user such as a developer, data administrator, or tester to access the system. User device 118 may be any appropriate device or machine, as described above with respect to client device 102. The environment also includes a development server 120, which functions similarly to application server 108, but typically traverses the code during development and testing before the code is distributed and executed on the production side and is accessible to external users, for example. In some embodiments, an application application server may function as a development server, and separate production and test storage may not be used.
[0023] Data stores of data plane 110 may include a number of separate data tables, databases, or other data storage mechanisms and means for storing data in relation to a particular aspect. For example, the illustrated data plane includes mechanisms for storing production data 112 and user information 116, which can be used to serve content to the production side. The data plane is also shown to include a mechanism for storing test data 114, which can be used with user information for the test side. It should be understood that there may be many other aspects that may need to be stored in a data store, such as for page image information and correct access information, which may be stored in any of the mechanisms listed above as appropriate or in additional mechanisms. in data plane 110. Data plane 110 is operable, through logic associated therewith, to receive instructions from application server 108 or development server 120, and obtain, update, or otherwise process the data, instructions, or other such information in response thereto. In one example, a user might submit a search request for a certain type of item. In that case, plan components can access user information to verify the user's identity, and access catalog detail information to obtain information about items of that type. Information can then be returned to the user, such as in a listing of results on a web page that the user is able to view through a browser on the user's device 102. Information for a particular item of interest can be viewed in a dedicated page or browser window.
[0024] Each server will typically include an operating system that provides executable program instructions for the administration and general operation of that server, and will typically include a computer-readable medium storing instructions that, when executed by a server processor, allow the server to perform its intended functions. Suitable implementations for the operating system and general functionality of the servers are known or are commercially available, and are readily implemented by persons having ordinary skills in the art, particularly in light of the disclosure herein.
[0025] The environment in a modality is a distributed computing environment using several computer systems and components that are interconnected through communication links, using one or more computer networks or direct connections. However, it will be appreciated by those of ordinary skill in the art that such a system could operate equally well in a system having fewer or more components than are illustrated in FIG. 1. Thus, the representation of system 100 in FIG. 1 should be taken to be illustrative in nature and not limiting the scope of disclosure.
[0026] An environment such as that illustrated in FIG. 1 can be useful for multiple content providers or other such entities, where multiple hosts and multiple types of resources can be used to perform tasks such as serving content, authenticating users, allocating resources, or performing any one of a number of other such tasks. Some of these hosts can be configured to provide similar functionality, while other servers can be configured to perform at least some different functions. The electronic environment in such cases may include additional components and/or other arrangements, such as those illustrated in configuration 200 of FIG. 2, discussed in detail below.
[0027] Modal-compliant systems and methods provide at least one resource access path, or control plane, as part of the data environment or in a pathway between the user and the data plane, which allows users and applications access shared and/or dedicated resources, while allowing consumers, administrators, or other authorized users to allocate resources to multiple users, clients, or applications and ensure adherence to those allocations. Such functionality allows a user to perform tasks such as storing, processing, and querying relational datasets in a cloud without worrying about latency degradation or other such issues due to other users sharing the resource. Such functionality also allows guest users to gain access to resources to perform any appropriate functionality, such as rendering and/or serving streaming media or performing any of a number of other such operations. While this example is discussed in relation to the internet, web services, and internet-based technology, it should be understood that aspects of the various modalities can be used with any appropriate features or services available or offered over a network in an environment electronic. Furthermore, while several examples are presented in relation to shared disk access, data storage, hosts, and peripheral devices, it should be understood that any appropriate resource can be used within the scope of the various modalities for any appropriate purpose, and any parameter appropriate may be monitored and used to adjust access to or use of such feature by any or all of its users.
[0028] A resource path or control plane 208 may be used in some environments to provide and/or manage access to various resources in the 232 data plane. In a cloud computing environment, this may correspond to a manager in cloud 210 or similar system that manages access to the various resources in the cloud. In one embodiment, a set of application programming interfaces (APIs) 220 or other such interfaces are provided that allow a user or consumer to make requests to access various resources. Once access is established, a resource is allocated, etc., a user can communicate directly with the resource to perform certain tasks in relation to that resource, such as storing or processing data. The user can use direct interfaces or APIs to communicate with data instances, hosts, or other resources once access is established, but use the control plane component(s) to gain access.
[0029] FIG. 2 illustrates an example of a configuration 200, such as may include a cloud computing management system, that may be used in accordance with an embodiment. In this example, a computing device 202 for an end user is shown as being capable of making connections across a network 206 to a control plane 208 (or other such access layer) to perform a task, such as obtaining access to a specific resource or resource type. While an end-user computing device and application are used for explanatory purposes, it should be understood that any appropriate user, application, service, device, component, or resource can access the interface(s) and components of the component. connection and data environment as appropriate in the various modalities. Furthermore, while certain components are grouped together in a data “plane”, it should be understood that this may refer to a real or virtual separation of at least some resources (eg hardware and/or software) used to provide the respective functionality. Furthermore, the control plane can be considered to be part of the data plane in certain modalities. While a single control plane is shown in this modality, there may be multiple instances of control or access to services or management components in other modality. A control plan can include any appropriate combination of hardware and/or software, such as at least one server configured with computer-executable instructions. The control plane may also include a set of APIs (or other such interfaces) to receive calls from web services or other such requests through network 206, which a web services layer 212 may evaluate or otherwise analyze for determine the steps or actions required to act or process the call. For example, a web service call may be received that includes a request to establish a connection to a data store to perform a query to a user. In this example, the web services layer can evaluate the request to determine the type of connection or access required, the appropriate type(s) of resource required, or other such aspects.
[0030] The control plan may include one or more resource allocation managers 210, each responsible for tasks such as validating the user or client associated with the request and obtaining or allocating access to the appropriate resource(s) (s). Such a system can handle multiple request types and establish multiple connection types. Such a system may also handle requests for various types of resources, such as specific graphics processors or other types of hardware or hardware functionality, and may provide access to the appropriate resource(s). Data plan components, or the cloud resource layer, can perform the tasks necessary to provision the resource. To access a data instance, for example, this might include tasks such as provisioning a datastore instance, allocating a persistent storage volume outside the instance, attaching the persistent storage volume to the datastore instance, and allocating and appending an IP address (derived from DNS mappings) or other address, port, interface, or identifier that the consumer may use to access or otherwise connect to the data instance. For tasks such as obtaining processing of an instruction using a particular type of hardware, for example, the components of the data plane, in conjunction with the control plane, can perform actions such as providing a device to a user and providing of shared and/or dedicated access to the resource over a period of time at a particular level of access to the resource. In this example, a user can be provided with the IP address and a port address to be used to access a resource. A user can then access the resource directly using the IP address and port, without having to access or traverse the 208 control plane.
[0031] The control plane 208 in this mode also includes at least one monitoring component 214. When a data instance or other resource is allocated, created, or otherwise made available in the data plane, information for the resource can become written to a data store accessible to the control plane, such as a monitoring data store 216. It should be understood that the monitoring data store may be a separate data store or a portion of another data store. A monitoring component 214 can access information in the monitoring data store 216 to determine information, such as past use of resources by various users, a current number or type of threads or resources being allocated to a user, and other information. of such use. A component monitor can also call data environment components to determine information such as the number of active connections for a given user in the data environment and usage aspects of each connection. A component monitor can constantly monitor the usage of each resource across a user, client, etc., having an allocation provided through the connection manager. A component monitoring can also access information stored in an administrative (“Admin”) or similar data store 216, which can store information, such as the general allocation granted to a user, regulating or limiting information for a user, permissions resource for a user, or any other such information that can be specified and/or updated by an administrator or other such user.
[0032] In an example where users request connections to multiple instances of data, each instance 222 in the data environment may include at least one data store 226 and a host manager component 228 for the machine providing access to the data store. A host manager in a modality is an application or software agent running on an instance and/or application server, such as a Tomcat or Java application server, programmed to manage tasks such as software distribution and data storage operations. , as well as monitoring a state of the data store and/or its instance. A host manager may be responsible for managing and/or performing tasks such as installing instances to a new repository, including installing logical volumes and file systems, installing database binaries and seeds, and starting or stopping the repository. A host manager can monitor the health of the data store, monitor the data store for error conditions such as I/O errors or data store errors, and can restart the data store if necessary. A host manager may also perform and/or manage the installation of software patches and updates to newer versions, for data storage and/or operating system. A host manager can also collect relevant metrics, such as might relate to CPU, memory, and I/O usage.
[0033] The resource manager 210 may periodically communicate with each host manager 228 to which a connection has been established, or with an administration server or other component of the resource environment, to determine status information, such as load, use, capacity, etc.
[0034] As discussed, once a resource is provided and a user is provided with an IP address derived from DNS mappings or other address or location, the user can communicate "directly" with components or resources of the plan of data 232 over the network using a Database Connectivity in Java (JDBC) or other such protocol to directly interact with that resource 222. In various embodiments, as discussed, the data plane takes the form of (or at least includes or is part of) a cloud computing environment, or a set of web services and resources that provide data storage and access across a “cloud” or dynamic network of hardware and/or software components. An IP address derived from DNS mappings is beneficial in a dynamic cloud environment, as availability or instance failures, for example, can be masked by programmatically remapping the IP address to any appropriate replacement instance for a use . A request received from a user 202 or application 204, for example, can be directed to a network address translation (NAT) router 224, or other appropriate component, which can direct the request to real resource 222 or corresponding host. to the mapped address of the request. Such an approach allows instances to be dynamically moved, updated, replicated, etc., without requiring the user or application to change the IP address or other address used to access the instance. In some cases, a resource 222, such as a data instance, can have at least one backup instance 230 or copy in persistent storage.
[0035] As discussed, a resource can be shared among multiple users, clients, applications, etc., either concurrently or at different times, with variations in access levels or allocation. When a user has dedicated access to a machine or resource, the user may also have native or “scratch” access to the resource for a period of time, depending on the type of access required, and other such factors. Providing this level of access to a feature comes with potential risks for a provider of the feature, as a user having native access to the device may have the ability to modify firmware or other configuration information for the feature, which can affect capacity from a subsequent user using the resource without first re-imaging or otherwise checking the state of the resource.
[0036] Several modalities allow a provider to grant a user or consumer substantially complete access to a hardware resource with a reasonable level of security. Such native-level access to remote hardware can be provided to resources such as servers, hosts, and cluster instances, for example. For resources such as cluster instances, consumers may have native access to a subset of the hardware resources, such as may include peripheral devices connected using a component such as a Peripheral Component Interconnect (PCI) bus. These peripheral devices can include network interface cards (NICs), graphics processing units (GPUs), and similar devices that would often be virtualized in today's cloud environment. In some cases, a consumer may have full access to an entire machine, or groups of machines, including any or all devices built into it. For a group of machines such as a rack of servers, a user may be granted substantially full access to the entire rack, including any switches or other devices or components provided as part of the rack.
[0037] Certain providers present such hardware resources as a virtualized abstraction, so that management of physical hardware can occur in a "more reliable" execution context, and can provide additional benefits, such as the ability to migrate consumers to different resources without interrupting execution and, since consumers or “guests” are not tied to specific hardware, the ability of vendors to compete to provide the best value of utility computing for price. Also, smaller and simpler guest instance images can be used, as the guests do not need a multitude of hardware-specific drivers. Such virtualization can come with potentially significant costs, however, as virtualization can incur an order of magnitude performance penalty for hardware that does not include native acceleration for virtualization, and virtualizing a particular hardware device can consume substantial non-relative resources that device (eg, a processor and/or memory used to virtualize a network interface). Also, virtualization support can lag years behind convenience availability of new hardware (eg, video cards), and certain appliance hardware is often too specific or “niche” to at some point ensure virtualization support convincing. There are potentially huge market opportunities in supporting high-margin niche appliances or being the first to market cloud support of new types of hardware. Providing such support through native access, however, can lead to several vulnerable aspects of the internal cloud, such as provision of technology, billing, resource utilization and balancing, and 2-layer network layout, for example, and it can violate threat models well beyond consumer requests.
[0038] Various modalities can provide "partial" or "substantially" full access to a resource, such as a host server, by providing users with native access to host hardware or specific devices, such as cards plugged into a peripheral control bus or via data from similar hardware. In certain modalities where specific performance levels are an issue, technology such as an input/output memory management unit (I/O MMU) can be used to “assign” peripheral devices to guest operating systems (for example , virtualization technology for directed I/O (Intel's VT-D)), effectively giving guests native access to only those peripheral devices. As should be evident to one skilled in the art, a guest operating system (OS) can refer to different systems in different modalities, such as a virtual machine hosting a functioning OS with at least partial non-virtualized access of which some hardware state or machine that the OS or hypervisor depends on including BIOS, configuration, etc., which is not under the hosting provider's administrative control. In other embodiments, the guest OS may refer to an OS that is not under the administrative control of the hosting provider in operation without full virtualization. In one modality, an MMU can logically connect to an I/O bus capable of direct memory access (DMA) (eg, a PCI bus) to main memory on a host, and can manage the mapping of I devices /O for physical addresses to regulate the flow of information from a guest to multiple PCI or similar devices. These devices can include, for example, graphics processing unit (GPU) coprocessors, high-performance NICs, disk controllers, or other “niche” coprocessing devices such as cryptographic cards or hardware codecs. In some instances, virtualization or other such technology can be used to provide a level of separation between guests and host machines from the host hardware (eg CPU, memory, etc.), with native access potentially being available to specific devices on a given host. In other embodiments, native access can be provided to any hardware included in, or available to, a specific host.
[0039] One of the main issues in providing consumers with native access to specific hardware is that consumers may have the ability to modify privileged configuration or BIOS (basic I/O system) options, or other firmware images on host hardware . These changes may persist throughout a physical system reboot, so the hardware may not return to the same state the hardware was in before the consumer is granted access to the host or its device(s). In the case of dynamically configurable options for a virtual machine monitor (VMM) managed by a Ring 1 hypervisor, for example, changes would generally not persist across reboot, but could persist across operating system instantiations guests in a virtualized environment (for example, integrated circuitry options to support IOMMU technology). This ability for a consumer to modify options or firmware that would otherwise be immutable can have serious security implications. For example, malicious software (eg Trojans or viruses) can be inserted into firmware for various devices. Even if the firmware changes do not involve intentionally malicious programming, however, the changes could still be unintentionally harmful by causing performance and/or compatibility issues. The firmware swap process can potentially physically destroy the hardware irreparably (also known as hardware bricking). Certain technologies have been developed that can address at least some of these challenges, particularly for motherboard firmware or integrated circuitry configurations. These technologies include, for example, Trusted Platform Module (TPM), Intel's LaGrande Technology (LT), Measured Boot Technology, Trusted Boot Technology, Dynamic Root of Trust (DRTM) and Static Root of Trust (SRTM) technology ). None of these solutions, however, are known to address many issues specific to device firmware, complete hosts, and other such hardware aspects.
[0040] Systems and methods complying with various modalities may prevent and/or monitor access and/or manipulation of firmware images or configuration information by guests in a cloud or similar electronic environment. In certain embodiments, a consumer can be provided with dedicated guest access to a hardware resource for any desired period of time, such as a matter of hours or even minutes. FIG. 3 illustrates an example of a configuration 300 that can be made to provide such native access to a consumer in accordance with a modality. This example will be discussed in relation to granting a user access to a peripheral device on a host machine using conventional PCI-based technology, but it should be understood that this is merely an example and that approaches within the scope of the various modalities can be used with any appropriate hardware (including based on different bus technologies or with greater or lesser degrees of system integration within individual components or “chips”), software, and protocols currently used or subsequently developed for such purposes.
[0041] This exemplary configuration 300 includes a set of host devices 302, such as servers or similar devices, which may each have a series of network ports 304. Some of these ports may function as "production" ports that connect each host to at least one network switch 306 capable of processing and routing network traffic to/from each device. In some embodiments, the network switch can be a "smart" network switch, while in other embodiments, segregation can occur at a higher level in the network than the first floor of switches. In an example data center, there might be one smart switch for each rack of 308 servers, for example. At least one of these network ports 304 can host traffic to a guest operating system, where the guest is effectively operating "on top of" at least one central processing unit (CPU) 310 on the partitioned or allocated host device (e.g. , server) 302 that has access to that production network port. The host device 302 can also have at least one terminal port 312 and a terminal controller 314, which can connect to a separate terminal network 316. This “terminal network” can also be implemented using the same network technology as the “production network” such as Ethernet technology. In some embodiments, at least some of these ports can be merged but logically separated (eg, multiplexed on the same physical port). Each host device could also have one or more dedicated power supply units (PSUs) 318, which can be accessed by the terminal controller and/or the main CPU, through which the machine can be powered off through the host CPU or a device. on the network, for example. The power supply for each server in a rack can be connected to a rack power distribution unit (PDU) 320, which can be connected by a larger power cable to one or more data center PDUs 322 each. which can support multiple rack PDUs. In some cases, hosts 302 can be turned on and off by running a line to the terminal controller from the rack PDU with relays or other such components to power cycle each device.
[0042] At least one router 324 can connect host devices to one or more provisioning systems 326, and the switch and/or router can manage access to those provisioning systems. In some embodiments, network traffic within a rack is aggregated in order to minimize the number of cables leaving each rack. In some embodiments, a capability such as a preboot execution environment (PXE) exists on a host machine 302 at production network port 304, so that power can be circulated using the terminal and when the machine stops. boots, the PXE code can run on the network port. PXE access could also be enabled or disabled depending on the type of reset that was authorized. For example, reboots could be allowed from local images on the host to consumer initiated reboots, but PXE access could be disabled upstream. When switch 306 is configured to connect a host machine 302 to the provisioning systems, the PXE can connect the device to the provisioning systems and boot the machine to a RAM disk (random access memory) or other block of storage, for example , which enables control operations, such as the firmware change process or provision of a new vendor image. A RAM disk with drivers specialized in one modality can be used to boot and/or run an unknown or unreliable image, which might otherwise not be able to boot on a specific machine. The provision of images can thus be received over the network to the PXE, which contains the provision code or firmware exchange process code. Once provisioning is complete, authorized consumer networks 328 can interface with devices 302 through switch 306. Control and provisioning systems can control the switch in real time without humans involved, since automatic switching of that path can be with based on provision events and external coordination, for example. Coordination may be provided and/or managed by an external system, such as a cloud manager system and database 330, or other control plan or control system as discussed elsewhere here, which can instruct the( s) 326 provisioning system(s), 316 terminal network, and rack components to perform certain actions. Cloud manager 330 can include one or more workflow systems that work with a central database, in one modality, to perform various aspects of resource management.
[0043] In an environment such as a cloud computing environment where different physical servers can be used for host consumers at different times, it may be desirable to provide a level of abstraction for a consumer user or network to avoid dependencies on resource allocations that can change over time. The presentation of virtual network equipment, such as consumer network routers and consumer network firewalls, can also be achieved using overlay network technology. For example, a consumer virtual LAN or other virtual network between multiple computing nodes may be provided in at least some embodiments via an overlay network across one or more intermediate physical networks separating the multiple computing nodes. The overlay network can be implemented in various ways in various modalities, such as by encapsulating communications and embedding virtual network address information for a virtual network into a larger physical network address space used for a network protocol. one or more intermediate physical networks.
[0044] This allows consumers to use a standardized address space to address resources on the consumer's network. By using a standardized address space, a consumer can create an overlay or “virtual” network that can use common base addresses, subnets, etc., without the restrictions that the substrate network places on the physical address space.
[0045] Using virtualization, a number of virtual machine instances can be generated that appear and function to a user as being a part of the consumer's network, those that are mapped to real servers or other physical resources in a separate or remote cloud , network, etc. As discussed, using a standardized address space may require constructing and maintaining a mapping between the physical substrate addresses and the overlapping virtual addresses that are used for the consumer address space. In some existing approaches, a central processing unit operating on a host device can control the mapping of physical and virtual addresses so that a request received from a consumer can be directed to the appropriate resource. This can take the form of packet data encapsulation and de-encapsulation, for example, where the physical address and/or header information can "coexist" at various times with the virtual address and/or header information, so that a packet can be addressed to the virtual address by a source on the consumer's network, but can be properly routed to the appropriate physical address by adding the physical header information when in cloud or remote network infrastructure.
[0046] For example, FIG. 4 illustrates an example where a packet 400 received from a consumer or "overlay" network is encapsulated in order to be routed within a physical substrate network in which the virtual cloud environment is hosted, in accordance with an modality . In this example, the received consumer packet 400 includes three main parts: a virtual address 402 (such as a "virtual IP address" relevant to the consumer overlay network, denoted here as "IPv"), a 404 protocol header (such as an original transmission control protocol header as found in the internet protocol suite, denoted here as “TCPo”), and a portion of “payload” or data 406. The virtual IP address can be an address only relevant to the consumer or overlay network. In order to properly route the packet to the intended destination host, that packet may be encapsulated to include an "external" data structure or frame that can route the packet within the substrate or cloud network or other such pool of resources. In this example, the encapsulation process is shown as producing a “substrate” packet or datagram 410, which includes the IPV, TCPo, and the payload of the original consumer packet, but has additional “header” information added to it, including here a “real” or physical address 412 (such as the IP or “IPR” address within the cloud substrate network) and a control header 414 (such as a protocol header useful via the control plane to process and/or route the package). Without the addition of any of this “real” information, the routers and other such components that host the cloud infrastructure would generally not be able to properly route packets to the appropriate destination(s), once the Consumer routing information (eg modalized by 402) is only meaningful to the consumer overlay network and not the physical network infrastructure to which the cloud host resources are connected. In some embodiments, any consumer packet being received for a device in the cloud can be encapsulated to include that physical routing information for use within the cloud. Since the first device to receive a packet in the cloud can be considered to be on the “edge” of the cloud, these devices will be referred to here as “edge” devices. An "edge" device as used herein may refer to any device in hardware and/or software capable of receiving a packet of information from outside the cloud, and/or capable of transmitting a packet of information from within the cloud. a cloud. The encapsulation process can take place on any appropriate edge device in some modalities, while in other modalities the edge devices can route packets to an encapsulating component or other device capable of encapsulating or de-encapsulating the packets. As should be understood, when a packet is to be transmitted back to the consumer's network, or otherwise transmitted outside the cloud, a “de-encapsulation” process can be performed, in which the IPR 412 and a control header 414 are removed and the packet can be routed using the virtual address space information to the consumer's network. For purposes of simplicity, the encapsulation process will be discussed in relation to the various embodiments, but it should be understood that an unencapsulation process can also be performed using such components and processes in accordance with the various embodiments.
[0047] Certain conventional approaches perform a level of encapsulation in hardware such as host devices and servers. In these approaches, a central processor can perform the encapsulation procedure in order to route incoming packets to a network port, network interface card (NIC), or similar device. The encapsulation process is generally not exposed to the user. In some embodiments, the driver for the NIC would be directly accessible by the processor, so the processor can access a mapping engine or distributed mapping service to map physical substrate packets to virtual overlay packets, and vice versa, before route packets to or from consumer networks via the NIC. In some cases, mapping information can be distributed from a centralized service to each appropriate node across the cloud.
[0048] As discussed, however, a resource provider may want the ability to provide users or consumers with substantially full native access, or “from scratch” access, to a hardware resource, such as a host machine. If the mapping is managed by an application running on a host machine's CPU, for example, then that mapping can potentially be accessed through a user or guest operating system (OS) running on the host machine. Such access could potentially compromise the mapping service, and could allow a guest operating system to redirect packets, reject packets, or otherwise impact packet processing on the cloud network. Furthermore, such functionality could be compromised so that packets can be sent to unintended locations outside the cloud. Other potential problems include “packet masking,” in which a host sends packets that appear to originate from a different host or location. This is often used to obfuscate where adversarial attacks are coming from, and can also be the basis of “ACK-based” Denial of Service (DoS) attacks, where acknowledgment packets that are part of standard network protocols are sent to hosts that never initiated transmissions, etc. Several other potential issues arise when the guest OS or CPU potentially has access to mapping and/or encapsulation functionality.
[0049] Correspondingly, systems and methods conforming to various modalities can provide access resources substantially "from scratch" by multiple users, while performing operations such as encapsulation, de-encapsulation, and session-state firewalling operations using components that are not exposed to the consumer, guest OS, CPU on a provisioned host machine, or other sources of potential manipulation. FIG. 5 illustrates an example of a configuration 500 that can be used to perform packet processing and provide other network functions in accordance with various embodiments. In this example, packets are encapsulated "upstream" from consumer accessible host resources, here at the network card level, such as just before a packet is in frames for physical interconnect transmission (eg, Ethernet framing) . In this example, it can be seen that the offload device 506 has an external port 508 that can communicate with components such as the cloud manager 504 and a mapping service 502. The external port 508 can allow these components to communicate with the CPU independent offload device 514 on the host machine 516, or any guest image 518 or the guest OS provided on the host machine. Using such an approach, any packet transmitted to or from the cloud can be processed independent of the portions accessible to the guest, so the mapping is not accessible to, or modifiable by, the user. In that example, the offload device may have memory 510 and a processing device 512 capable of performing at least basic mapping, encapsulation, de-encapsulation, and/or such similar functions. This will generally be referred to here as “discharge device-based” packaging, although it should be understood that other peripheral devices or hardware components may perform similar functionality, and that the functionality is not limited to packaging, but may also include other functions, such as uncoating, firewalling, etc. An offload device can function as an embedded system on the host machine that is not exposed to the user or guest operating system. In cases where the user may want native access to at least some of the offload device's functionality, the offload device may only have certain portions of memory mapped to the guest OS so that only some functionality can be accessed. In some embodiments, this can take the form of a virtual dump device image, where the guest OS can discover and/or use portions of the dump device, but cannot access portions used to secure actions such as encapsulation.
[0050] The offload device-based encapsulation functionality may be provided on a per host basis, or at least for those host machines capable of receiving and/or transmitting packets, and/or capable of having a provider image provided in the same. In such cases, the cloud manager 504 or a similar component or system can manage the distribution of mapping information to the various hosts and/or nodes, as well as other such aspects and configuration information useful for such processes. In such cases, the cloud manager may communicate with a offload device 506 through external port 508 to update configuration information, firmware, or other information useful to perform encapsulation and such similar actions. Processes for updating configuration information via an external channel are disclosed in co-pending US Patent Application No. 12/554,690, filed September 4, 2009, [ATTORNEY DOCUMENT 026014-010700US], which is incorporated herein by reference. Using such an approach, the firmware and/or configuration information for the offload device can be updated to perform the desired functionality, as well as to communicate with the 502 mapping service or other appropriate component(s) ) as necessary. The configuration can be updated periodically as it can be managed by the cloud manager and/or mapping system(s), such as to send large payloads or otherwise adjust the functionality of the offload device.
[0051] In some embodiments, encapsulation and similar processes can be performed on other components that are not exposed to the user, such as a 520 smart switch configured to route messages to and from a 506 offload device and/or network port 520 of a host machine 516. Such a switch may include a processor 522 operable to perform operations such as packet encapsulation, whereby the switch may process and route packets to the appropriate addresses in virtual address space and/ or physical. In such cases, the host machine can be considered (from an address space perspective) as being outside the cloud, or trusted environment, whereby the switch can act as an edge device and modify packets received from from the virtual address space of the host machine (and client networks) to the physical address space of cloud resources. Various other components can also be used, such as routers or dedicated edge devices, within the scope of the various modalities.
[0052] One of the limitations in many conventional systems is that the physical transmission path or "cable" can only allow relatively small packets of information, such as 1.5KB or 9KB packets. The use of smaller packages is not strictly a physical consideration, but it is also a result of protocol definition and historical reasons. For example, in modern networks where most or all links are switched and transmission rates are high, this limitation could be increased by orders of magnitude without increasing collisions intolerably. Although a physical network interface, such as a dump device, can only transmit or receive 1.5KB or 9KB packets, it is desirable, in at least some embodiments, to transmit larger packets from the DOM network stack -U to DOM-0 and then to dump device, and have the dump device segment the largest packet into multiple 1.5KB or 9KB packets. Many convenience download devices support advanced functionality such as segmentation download to address this requirement. A offload device with segmentation offload capabilities can be configured to receive and/or load relatively large packets, and segments or frames those larger packets into smaller packets or Ethernet frames that comply with the 1.5KB, 9KB, or restriction another size restriction. Devices receiving these packets can be configured to rearrange larger packets based on the plurality of smaller packets.
[0053] Many offload devices provide advanced features such as TCP segmentation offload that can assist with high speed networking. Systems and methods conforming to various embodiments can take advantage of such features to provide "virtual" networking, such as where a consumer has access to a host device seated between a consumer address space and a provider network address space. Typically, the segmentation offload functionality works only with well-known level four (“L4”) protocols such as TCP. When packets are encapsulated, as described in the previous paragraph with respect to FIG. 4, L4 protocol is changed to something other than TCP. Thus, the segmentation download features in the download device are not able to work on such encapsulated packets. As used in the technique to describe layers between physical hardware ("level one") and an application running on that hardware ("level seven"), level four refers to a level of "protocol", which, in the case of Internet protocols, may refer to protocols such as Transmission Control Protocol (TCP) and User Datagram Protocol (UDP). Receive-side TCP segment processing assumes that the TCP segment payload is completely consumer data (or other such data). Thus, on the transmit side, encapsulation related metadata cannot be added to the L4 payload in order to retain the original L4 header, as adding metadata would cause the receiving side to corrupt the packet payload with metadata of encapsulation/decapsulation.
[0054] Another potential problem with the existence of encapsulation and/or overlay network implementations is that headers often do not include physical port information, which is used by conventional hardware devices for purposes such as routing and load balancing.
[0055] Several modalities can use false TCP header with original port numbers or, in some cases, false ones, where the header is extended following the established protocol rules (for example, TCP options) and the encapsulation/de-encapsulation information is traversed in the protocol extension. A “false” TCP header, for example, can include any port information appropriate to the convention in addition to any appropriate TCP-related information. By including this false port information, conventional routers and other such devices can achieve improved load distribution, as many load distribution decisions are based on conventional hardware devices at least in part through the port specified in the header. A router or offload device can see an IP address and TCP information, for example, and can process the packet as a standard packet. Such an approach can also be advantageous as it can be implemented primarily in software using conventional hardware devices and networks.
[0056] A protocol can also be used that does not change the level four payload (in the network stack, as discussed above). Because the original packet received from a user can include the payload (here, a level four payload), along with a virtual IP address (at level three in the network stack) and an original TCP header (in level four). Using an encapsulation approach, as discussed earlier, a controlling host can append a true address, such as IPR, and a fake TCP, TCPF, (or UDPF, for example) header for use in packet (or frame) routing on the physical or secure network. For the packet after encapsulation, the original virtual IP address, TCP (or UDP, etc.), and payload information now effectively forms the level four payload, with IPR forming the level three addresses and TCPF forming the level four protocol header. Since packets have original or fake port numbers, such a format can also resolve issues such as the aforementioned router ECMP scatter issue. A conventional NIC or similar device, however, will not know how to properly divide a 64K or similar packet according to the encapsulated frame, since the NIC will not be able to properly interpret the information now contained within the level four payload. Also, as discussed, the level four payload was changed through the inclusion of IPv and TCPO information.
[0057] Various modalities can, on the contrary, take advantage of a slightly modified protocol format to handle the encapsulated packets. Conventional protocols provide extra space at the end of a TCP header, which typically allows for what are referred to as "TCP options" or "TCP add-ons". These TCP options allow the TCP protocol to be expanded to include additional features. In some embodiments, the TCP packet will effectively be extended by about 24 bytes, with the additional information always declared as a TCP choice. As it should be understood, packets can be extended by different amounts in different modalities and/or implementations, and a 24-byte extension is just one example. The fake TCP header thus can include the original TCP information, plus the control header information. Information for the virtual IP address can also be included in that TCP choice space. So, instead of adding the actual headers during payload encapsulation and modification, IPv and TCPO information can be included in the TCP fake TCP section options so that the payload or data portion is unchanged.
[0058] In an exemplary process for managing packet information regarding a virtualized environment, a packet is received that includes virtual address information. If received to a host device or other machine to which the user has substantially full access, the packet is routed to one or more upstream devices or components of the user-controllable hardware, so that the user is unable to modify the routing and other such processing. Packet passed between components, such as from guest to DOM-0, can be up to 64KB in size in some modalities, and thus may require segmentation. The mapping information for the packet can be determined, such as by contacting a mapping service to determine the physical address information that corresponds to the virtual address information. Address information can be added to the received message, such as a header (such as an IPR section), where the address information corresponds to the physical address to which the packet is to be directed. Virtual address information can be added to a protocol header, such as a TCP header, for the packet, without modifying the payload, so that the packet can still be routed, segmented, and otherwise processed through the convenience hardware. The packet is transmitted to the offload device, which can segment the packets using TCP segmentation offload functionality and transmit the resulting packets over the cable, and then to the final destination. As should be evident, similar functionality can be used to process packets received from a physical address space, where mapping information is determined for the packet and virtual address information is added to the packet. Where the virtual mapping information does not specify a port, a “false” port can be used, which allows the packet to be processed on its way to the virtual destination, such as to allow for load balancing or similar functionality.
[0059] In an example of a similar process for managing packet information in relation to a virtualized environment, an Ethernet frame is received for a physical network interface (eg a NIC), where the frame includes address information physicist. Segments with information, such as IPR and TCPF, can be coalesced in some modalities to generate one or more larger segments, which improves performance. This can also be done through convenience NICs that support Receive-Side Coalescence, since the packet format follows all TCP format rules and the TCP payload is exactly the same as the payload packet. of the consumer. The offload device (or other such device) is upstream of the user-controllable hardware, so that the user is unable to modify routing and other such processing. Virtual address information can be extracted from the protocol header, such as a TCP header, into the payload, after removing the header and footer framing information, for example. The virtual address information can be used to assemble a header for the data packet, extracted from the received Ethernet frame. The packet can then be processed, such as by transmitting the packet to a destination in the virtual address space. As should be evident, similar functionality can be used to process Ethernet frames received from a virtual address space, where virtual address information is extracted from the header for the packet.
[0060] Simply extending the TCP header may not be desirable in some embodiments, however, as if each received packet is 1.5K, and 24 bytes of information are added to each of those packets, then the packets would now be each one of more than 1.5K production limit and each would need to be split into two packets, which can lead to an unwanted amount of excess and additional traffic. It, therefore, may be desirable in at least some modalities to utilize this additional information while not significantly increasing the surplus.
[0061] Several modalities take advantage of the fact that information, such as IPv and TCPO information, is not required for every packet through segmentation, but can be determined through desegmentation. One approach, therefore, is to take the additional information for IPv and TCPO, etc. information (about 24 bytes in an example) and create encoded information (about 120 bytes in an example), which, in one modality, it is approximately one to five instances of information in various modalities, although other lengths of encoded information may be used as well, as it may depend on the dispersion technique. The encoded information can be reconstructed using a sparse or similar mechanism so that the original information can be reconstructed from at least 24 bytes of sparse metadata, which could be obtained from one or more instances of the packet segment. So, instead of triggering 24 bytes for each packet segment, for example, the additional 120 or so bytes can be divided into appropriate number of pieces and can be strategically positioned along the payload, such as at boundaries where the data will be targeted. For example, an offload device or similar device may know that data will be automatically segmented based on size at certain locations (including the additional 50 bytes). Once these segmentation locations are known, the offload device can insert instances of the additional information into these segment lines (or otherwise within different segments) so that at least five of the 1.5K packets ( or any other appropriate number of an appropriate size) will have information for IPv and TCPO stored in them, but each packet will not include all 10 bytes of additional information.
[0062] When packets are received, a desegmentation process can occur as with conventional systems. When the 1.5K segments are assembled into the 64K payload, or during the desegmentation process, the pieces of information can be used to reconstruct the IPv and TCPO etc information. An advantage to using a process of dispersing and distributing information among the various packets, for example, is that IPv and TCPO information can be reconstructed even if 1.5K packets are lost, as long as at least two segments with the portions of information are received. The full payload may not be able to be reconstructed, but at least the header information can be reconstructed. Furthermore, the receiving device can simply request those 1.5K segments (eg Ethernet frames) that are not received, since the header information can be reconstructed, and therefore does not need to request the payload resend full useful. Such an approach can have much smaller jitter variation, as there will often be no need to resend large packets, which could result in large variations in performance. In the case of video traffic, for example, as long as the lost data is not significant, the lost traffic can be neglected and thus need not be required in at least some modalities. This is an advantage of being able to receive partial segments successfully.
[0063] In an exemplary process for processing packets in a virtualized environment, a packet is received from a consumer address space, which includes virtual address information. As discussed, the initial packet received from the user can be a 64K packet with IPv and TCPO information. The packet may be received or directed to a controlling host or other secure component such that it is less partially inaccessible to a user of a partitioned consumer device. Virtual address information can be translated into a true address using the secure component, such as by contacting a mapping service as discussed above. The TCP header (or other protocol header) can be updated if desired, but additional information such as IPv and TCPO information can instead be inserted into the data. When adding IPv and TCPO information to data, that “virtualization” information can be dispersed or otherwise divided into multiple chunks. If not yet determined, a secure device can discover segmentation boundaries for the transmission path, and boundaries for user segment payload can be determined. Portions of the virtualization information can be positioned adjacent to, or positioned relative to, segmentation boundaries in central packets of the payload. The “new” packet or frame can then be passed on to the offload device or other secure device such, for example, that it can automatically segment the packet into a set of packets of the given size, such as 1.5K packets, with the number of segments depending at least in part on the size of the total package. The IP and TCP header can be replicated for each packet, with potentially some small changes to compensate for the total change in size, using offload device segmentation offload processes or other such device. Packets can then be transmitted to the destination.
[0064] A similar process can be used to process packets for a virtualized environment, in which a set of Ethernet frames is received, at least some of the Ethernet frames including "virtualization" information that has been scattered or otherwise split into multiple servings. Virtualization information can be extracted from the underlying segment of each frame that includes a portion of the virtualization information in the associated payload. The virtualization information (eg header data) is reassembled, provided a sufficient number of frames including the virtualization information are received, and the received packets can be desegmented to the extent possible. If not all frames are received, but the header data is capable of being reassembled, a request for only the missing segments can be sent.
[0065] When at least most of the packets are finally received at a destination, or the device along the way to the destination, the device may attempt to desegment or reassemble these packets into at least one larger segment, if not the Full 64K or other package. Since two packets (or a smaller number of packets that were originally generated during segmentation where the number of packets needed is determined using the specific dispersion technique) with additional header information in the payload is received, in at least some modalities, these packets can be used to reconstruct the header data and desegment the packets, replacing the true address and protocol information with information for the client or virtual network, whereby larger mounted segments can be passed on to the client or another destination. In some modalities, desegmentation can take place on a downloading device or similar device, while in other modalities, desegmentation can take place using the guest operating system on a receiving device, etc. Furthermore, several steps of the above process can be performed in any other order, or in parallel, and fewer alternative or additional steps are possible within the scope of the various modalities.
[0066] Using virtualization, a number of virtual machine instances can be generated, which appear and function to a user as being a part of the consumer's network, but which are mapped to real servers or other physical resources in a remote cloud or separate, network, etc. As discussed, using a standardized address space may require constructing and maintaining a mapping between the physical substrate addresses and the overlapping virtual addresses that are used for the consumer address space. In some existing approaches, a central processing unit operating on a host device can control the mapping of physical and virtual addresses so that a request received from a consumer can be directed to the appropriate resource. This can take the form of packet data encapsulation and de-encapsulation, for example, where the physical address and/or header information can "coexist" at various times with the virtual address and/or header information, so that a packet can be addressed to the virtual address through a source on the consumer's network, but it can be properly routed to the appropriate physical address.
[0067] A framework can be implemented through conventional or other network components, such as convenience NIC devices, which can allow these components to support multiple protocols, such as a variety of different standard and proprietary protocols. These commodity devices can then provide the enhanced performance and other advantages used for the conventional protocols of these devices, regardless of the specific consumer format of the packets. A NIC vendor, for example, can implement a framework that allows the NIC to be used by a consumer with any compatible protocol, without any customization or need for special hardware.
[0068] In one example, a offload device in a network environment can process TCP segments. The consumer network may use packets of a size (eg 64K) that cannot typically be passed from the offload device out into the network, as the offload device may only be able to transmit network packets in the order of 8K or 9K in size for example (depending on network configuration and other such issues). As discussed above, there are technologies that allow larger packets to be segmented at the offload device into multiple Ethernet frames of the appropriate size (eg 1.5K or 9K, etc.). For example, TCP Segment Offload (TSO) and Receive-Side Coalescence (RSC) can be used for the ingress and egress endpoints, respectively, to increase network throughput performance by allowing the host to handle larger TCP segments (eg 64K in size). TSO is a technique for segmenting TCP packets into segments of the appropriate size for long network transmission, and RSC allows these segments to be reassembled on the other side of the network. In general, however, techniques such as TSO and RSC are not supported for packets encapsulated with proprietary protocol information, such as the additional header information illustrated in FIG. 4(b). For example, packets that are encapsulated using a proprietary format are typically larger than TCP packets and do not have TCP header information in advance, so the offload device will not recognize these encapsulated packets.
[0069] Through the implementation of an appropriate framework, however, an offload device or other appropriate network component can have the capability and specifications to map the encapsulated packet to something the component can understand as a TCP packet. Once an offload device recognizes the packet as a TCP packet, for example, the offload device can segment the packet, add the appropriate headers, and/or do any of the other things that an offload device would typically do to a conventional TCP packet. Even for packets encapsulated with any of a variety of different protocols, TSO and RSC can provide a significant improvement (eg, up to 80% performance amplification) as well as other well-established advantages. Furthermore, when implementing a framework, offload devices can not only be used with different protocols, but can also allow consumers to upgrade to newer versions or change protocols without having to purchase, upgrade to newer versions, or modify their existing hardware .
[0070] An opaque field can be used with an encapsulated packet to include any information used by the particular format or protocol of the consumer's network, such as GRE s or other such protocols. The opaque field in at least some modes is a TCP or UDP based header, or other such protocol header. In one example, the opaque header has a first set of information at an offset specified in the opaque field that indicates or identifies the particular format of the segment or package. For example, the information can be a two-byte field that includes a heat corresponding to a particular format. The network hardware may contain, or have access to, a mapping of values from the first offset value and corresponding formats in order to determine, from the value of the first set of information, the appropriate packet format.
[0071] In this example, the opaque field also includes a second information field at a second offset specified in the opaque field. This second field can be of an appropriate length, such as two bytes, and can include a value that specifies a flow identifier, or an identifier for a specific traffic flow, as can be useful for desegmentation. In some embodiments, this field can identify a single TCP stream (or other streams such as a UDP stream) along with a regular 5-tuple when performing an ISO or RSC operation on a particular format packet.
[0072] These examples may correspond to an environment for a specific protocol, for example, where the header has information such as the virtual network to which the packet belongs, the virtual machine from which the packet originated, and/or the machine to which the package is heading. This information will not change between packets within a common TCP chain. Slot IDs, or virtual machine identifiers, can be used for connection information since, in a virtualized network environment, for example, there might be two different virtual machines on the same physical host that belong to two different virtual networks. Those virtual machines could have exactly the same IP address, and could potentially communicate with someone who coincidentally has the same port and IP address. From a TCP point of view, 5-tuple might be exactly the same. Other information, such as source IP and destination IP, source port and target port, etc., can also be exactly the same. So, from a TCP standpoint, the connections appear to be the same connection, but could actually be on two different private networks. The use of slot IDs can uniquely separate these situations. For other protocols, values other than virtual machine identifiers can be used as would be evident.
[0073] In one example, an encapsulated packet is received for a offload device. The offload device, using the framework specifications, can analyze the packet to identify that the packet is encapsulated and should be handled differently than a conventional TCP or UDP packet. In one example, an encapsulated packet includes internal and external IP headers. The encapsulated packet also has an opaque field (which may appear as part of the payload), which can be used for protocol-specific information. The length of the opaque field, and the information contained in it, may vary between modalities. In order to identify the packet as being encapsulated, the external IP header can contain pre-configured protocol information. Furthermore, the packet can contain at least a two-byte field in the opaque field (although other sizes and locations can be used within the scope of other modalities as well). The two-byte field can be a pre-set distance from the start of the opaque field, and the value of the two-byte field can also be pre-set. The combination of the protocol information in the external IP header and the format information in the two-byte field of the opaque field can allow the offload device or network component to recognize that the packet is encapsulated, as well as the format of the encapsulation. Since the download device does not otherwise see other information in the opaque header, the opaque header can include information specific to any particular protocol without affecting the download device's processing of the packet. The two bytes in the opaque header can identify a specific packet format, which can help determine rules or policies for packet processing. Based on this information in the external IP header and opaque field, the offload device can analyze each received packet to determine if the packet can be processed using conventional approaches or if the packet is an encapsulated packet and must be processed according to special rules specified by the framework.
[0074] During a TSO process, for example, segmentation of egress (eg, outbound) TCP segments can be performed using a standard algorithm on the TCP segment data starting at the internal IP header. The large encapsulated packet is segmented into a number of packets of a size that allows segments to be transmitted across the network. In order for the framework to work with stateless tunneling as well, the opaque field is literally copied into each of the resulting segmented TCP/IP packets, and positioned between the inner and outer IP headers. The external IP header is copied for each resulting packet and appropriate adjustments, such as a change to the “length” information, can be made using the same logic applied to the internal IP header. Furthermore, an IP ID can be generated, which is part of the IP header, along with a checksum for the IP header.
[0075] Similarly, during an RSC process, TCP streams of packets or segments having special protocol format information are defined by regular 5-tuple of TCP ports, the internal IP addresses, the protocol field of internal IP, and the internal L4 ports (eg TCP ports or UDP ports), as well as an additional two bytes at the preconfigured offset from the beginning of the opaque field. It should be understood that TCP streams of special format packets will not overlap with regular packet streams. Furthermore, it should be understood that terms such as "packages" are used entirely for purposes of simplicity of explanation, but in other locations or instance processes may involve objects more commonly referred to as segments or frames, and the common name for a single object may change between these and other terms at various points in the processes discussed here.
[0076] RSC is performed using a conventional algorithm on TCP packet data starting at the internal IP header. When coalescing related TCP packets, the opaque field from the first TCP packet can be copied to the resulting TCP segment between the inner IP header and the outer IP header. The outer IP header of the resulting TCP segment can be coalesced in the same way as the inner IP header is coalesced. If there are restrictions on IP flags (for example, "Do not fragment" or "More bit") that force ingress packets to be ineligible for RSC, the restrictions can be applied to IP flags in both internal and external.
[0077] RSC can maintain scatter containers (or other queues or temporary storage locations) for each connection for which packets are being received. When a TCP packet is received, the receiving device can then determine the connection to which the packet belongs, using information such as IP and TCP information as well as sequence number bits in the external TCP header, and can queue pack into the appropriate dispersion container. For containers where there are already packets, the network component can try to merge the segmented packets until the complete packet has been coalesced. Conventional criteria may apply, such as sending the coalesced packet to the operating system when the size reaches a certain threshold or packets are queued for a specific length or time range.
[0078] In at least some embodiments, however, the concept of a connection will differ from a connection for standard TCP packet processing. Instead of the conventional 5-tuple mentioned above, connections will be determined based on a 6-tuple, which includes the standard 5-tuple TCP connection information along with the new piece of connection information identified in the two bytes of the field. opaque. Once the network component discovers that the packet must be processed using the special rules, the component uses the 6-tuple instead of the 5-tuple to discover the connection information, and then runs the RSC process essentially the same as for conventional packages, to coalesce the packages, check the sequence number, etc.
[0079] Additionally, the RSC also, in many cases, needs to get rid of the opaque bits completely, but that of the packets being coalesced, such as the first packet received in some modalities. In some embodiments, RSC may not be performed when opaque fields do not match, so opaque fields from others will not be discarded at least until those packets can be processed otherwise. After a copy of the opaque bits is received and stored (at least temporarily stored or cached, for example), opaque bits from all other packets to be coalesced, which match the stored copy of the opaque field, can be discarded by the device. discharge. In addition, since the total length of the packet is changing during the merging of the offload device, they will have to make appropriate adjustments to the checksum, IP flags header, or other such information, both for the internal IP header and for External IP. In opaque fields, and elsewhere, byte count and other aspects could also be guaranteed. In addition to the two bytes (or n-bytes) of information used for identification, the expectation is that the rest of the opaque bits will be exactly the same for all packets within a particular TCP stream. In an example protocol, the opaque information could correspond to a specific network identifier. There might also be other information, such as a virtual machine identifier or slot ID, that would be the same for each packet in a TCP chain. In particular, n-bytes can identify the packet as corresponding to a particular virtual machine.
[0080] In many modalities, the framework relies on specific pre-configured values. For example, as discussed above, the framework can rely on a pre-configured length of the opaque field as well as an IP protocol value that identifies special or specific formats for incoming packets. The length of the opaque field, in some embodiments, corresponds to the length of the header for the special package format. The IP protocol value could be any identifier appropriate for a specific protocol. The framework can expect the offset of the n-byte field in the opaque field identifying the format to be preconfigured. In some modalities this might correspond to a specific port value.
[0081] Pre-configured values specific to the opaque field may vary for certain protocols. For example, the opaque field length for GRE support can be 16 bytes in one modality, with the IP protocol value identifying a specific packet or segment format set to a value such as 47. identifies a unique stream can be set to a value such as 10 to point to a part of a 'key' field or other such value.
[0082] In the case of an exemplary protocol, the length of the opaque field may match the length of a protocol-specific header, with a value such as 20. The IP protocol value that identifies packets or segments of a particular protocol may be set to the IANA protocol number for UDP, for example, with a value such as 17. The offset value in the opaque field that identifies packets or segments of a particular format can rely at least in part on the specific UDP port used , such as a UDP destination port with a value of 2. The field value that identifies a single stream when performing TSO or RSC can specify the source slot and target slot IDs along with the regular 5-tuple connection for identify a single TCP stream. It should be understood that similar approaches can be used to determine values for other protocols within the scope of the various modalities.
[0083] As mentioned above, a goal of an environment such as a cloud computing platform might be to provide each consumer with the illusion that a portion of the network infrastructure is dedicated to that consumer. In order to provide this illusion, the platform needs to provide certain levels of performance, such as being able to include high throughput, low latency, and low jitter network performance. While instability should generally always be low, the definition of low latency and high throughput for a given implementation depends on factors such as physical network equipment and product design, and will vary between instances. The illusion can also be provided in part by allowing the consumer to define a custom level two (L2) or level three (L3) network topology, with no addressing restrictions resulting from other consumer preferences. In certain environments, such as the Virtual Private Cloud (VPC) environment offered by Amazon.com, Inc. of Seattle, Washington, the option of a customizable L2 or L3 routable network is largely achieved through the implementation of sophisticated tunneling software of IP address. In at least some of these software implementations, however, it can be difficult to maintain high throughput, low latency, and low jitter network performance in a virtualized environment. The problem could be further exacerbated as current hardware trends continue with more cores, RAM, and virtual machines per host, placing an increasing burden on the networking subsystem. While gains can be made by optimizing the end-to-end software stack, it can be beneficial in at least some environments to provide hardware assistance in virtualizing network resources.
[0084] In order to satisfy at least some of the goals outlined above, hardware, such as various offload devices, may need to include several features. As used herein, "hardware-based" processing generally refers to any processing where a hardware device performs at least a part of that processing, or where a processing component presents itself as a physical device (eg, a NIC), but can actually be implemented as hardware and/or software. In some embodiments, hardware-based processing may be provided through a generic offload device or embedded system that appears to system components to be at least one hardware component. As an example, a generic unloading device can be used, which has an SR-IOV device in itself. A discussion of these features will be provided by giving a high-level overview of the proposed ingress and egress pathways, followed by details on individual stages that can be implemented in accordance with various modalities. For example, FIG. 6 illustrates an exemplary format 600 of such a virtual package. FIG. 7 illustrates a high-level overview of an exemplary offload hardware egress process 700 that can be used with such consumer packages from a consumer in a virtualized data center in accordance with at least one modality. As part of the egress process, an SR-IOV Virtual Role (VF) assigned to the consumer virtual machine receives an egress packet destined for consumer virtual network 702. In this initial state, the internal components 608, 610, 612 do package header 600 are present, while external components 602, 604, 606 and 614 are not present. One or more generic checks can be applied to 704 egress packets. These checks can include, for example, L2 and/or L3 source anti-tag as well as lock-in for all broadcast and non-IP (or ie for DHCP, ARP, etc. services). The offload device can perform a lookup against a pre-populated rule table 706, as it can based on an L2 target and an L3 target with a subnet mask, with a generic case being a subnet. IPV4 network “/32” that specifies a single target. Assuming a rule hit with a forwarding rule type, the rule can also specify an in-system memory pointer to the tunnel header that the offload device will include in the outgoing packet. At that point, the package may also include the initial external components 602, 604, 606. The offload device may perform one or more metric updates 708, which are discussed in more detail below.
[0085] Based at least in part on the rule match (or lack of a rule match), the offload device may determine an appropriate action to be taken 710. Actions may include, for example, to lock the domain of trusted root 712, drop packet 714, or forward packet with encapsulation and/or deformation 716. If the offload device decides to lock the packet to trusted domain 712, a driver callback may allow the trusted domain to perform processing to add-on software base. If the unloading device decides to leave package 714, no further processing will be done (in at least some arrangements). If the offload device instead decides to forward the packet 716, further processing may be required before the packet can be released to the physical network. In this example, the offload device performs the QoS and throttling action on packet 718, as described in more detail below. The discharge device may also construct and/or deform the final package that will be fed to the discharge motor 720. Outer package header components 602, 604, 606 can be added to the package. These can be retrieved through spreading and/or DMA assembling along with the packet bytes based on a previous rule match. The discharge device can then perform discharge(s) 720, including TSO if applicable. Packet header fields can be updated as needed including, but not necessarily limited to, internal and external IP length, internal or external TCP checksum (ie if IP protocol is TCP), MAC source of internal L2 and destination address, and internal L3 IP TTL, as discussed in more detail below.
[0086] FIG. 8 illustrates a similar high-level overview of an exemplary offload device 800 hardware-based ingress process for a consumer package in the virtualized data center that can be used in accordance with at least one modality. In this exemplary process 800, a packet is received at the dump device physical function 802. The dump device may construct a rule search key that will be constructed for subsequent rule processing 804, as discussed in more detail below. The offload device can then perform a lookup against a pre-populated rules table based on the derived search key 806. The offload device can perform various metric updates 808 as needed, and determine the appropriate action to be taken based on the less partly in a rule match (or rule mismatch) 810. In a first action, the offload device might decide to lock the packet to the trusted root domain 812. In that case, a driver callback might allow that the trusted domain performs additional processing to the software-based package. In another possible action, the unloading device may decide to leave the packet 814 so that no further processing of that packet will take place. As another possible action, the offload device may decide to forward the packet to an internal VF 816, such as with encapsulation and/or deformation. The VF ID (VM) can be specified in the forwarding rule. The discharge device can withdraw the outer casing header 602, 604, 606 from the pack 818. No inner deformation is required in this example, since all such deformation was previously performed in egress. Various other packet modifications can be performed as well, such as to reorder, split, or otherwise modify one or more packets or portions of packet data. At this stage, the package can be delivered to the Guest VM via Guest VF 820.
[0087] As mentioned, such an approach can provide both hardware-based and rule-based packet packet deformation and encapsulation. Such an approach allows consumer of multiple (and possibly overlapping) virtual networks to be overlaid on a unified L3 routable physical substrate. A common rules table can be used for both ingress and egress packet routes, the rules table being populated by the trusted root domain through software mechanisms in at least some modalities.
[0088] The following provides guidance on the size and performance of an exemplary rule table implementation that can be used conforming to various modalities. An exemplary rules table can be on the order of about 1,000 rule entries (shared between ingress and egress) per virtual machine running on the host. While in at least some embodiments it may be desirable to use the largest possible rule table size, there will in at least some cases be a limit on the rule table size imposed by the RAM device, as the primary cost of the larger table size will be higher RAM requirements on the offload device. As the number of VMs on a host increases, the number of rules can vary correspondingly. For example, if there are 128 VMs and 128 corresponding SR-IOV VFs, there would be 128,000 rule entries in at least one modality, although a number such as 32,000 or 16,000 might be sustainable. Rule entries in at least some modalities must be divisible across VFs as defined by the trusted root domain. For example, one VF could have 10 rule entries while another VF at 2,000 out of the total possible rule entries. The performance of rule table updates must also be fast enough so that it does not cause excessive queues in packet route processing. In some modes, the entire rule table can be modified in the order of every five seconds or so during normal operation.
[0089] An exemplary egress rules table can have a variety of different fields. In one example, a rules table has an internal L2 Target MAC (Matching Target) field. All egress rules can be matched on the internal L2 MAC address. This allows the consumer virtual network to be L2 only if desired (and to support protocols like RoCE that are not L3 aware). The table can also have an optional built-in IPV4/IPV6 destination with subnet mask field (matching target). Egress rules can optionally be matched on the target IP address/subnet. Using subnet rules allows multiple rules to be broken if desired. An optional internal L2 MAC source / target strain overrides field can be used as well. In order to support an arbitrary L3 topology, an ability to exchange both internal destination and source MAC addresses to support “ghost routers” can be supported. A VM might, for example, believe that it is on subnet A and is trying to send a packet to subnet B. Thus, the packet could have an L2 header as constructed by the guest VM, such as: Source address of L2 MAC: host 1 VF offload device MAC address (subnet A) L2 MAC destination address: subnet A path MAC address
[0090] At egress time, it may be desirable in at least some embodiments to be able to dynamically deform the inner L2 HEADER to look like the following example (so that when the packet is decapsulated in the target, the inner L2 HEADER looks like what is expected to be real router(s) between the 2 virtual machines): L2 MAC source address: B subnet path MAC address L2 MAC destination address: L2 MAC address Host 2 VF offload device MAC (Subnet B) An optional internal IP TTL decrement field can also be used. In order to support “ghost routers”, for example, the ability to optionally auto-decrement the internal IP TTL (if applicable) may be required. If the TTL reaches zero, the package must be locked to the trusted root partition.
[0091] The table can also have a field, such as for a pointer to an encapsulation bubble in system RAM. A bubble table can be stored in memory belonging to the trusted root partition. These memory addresses can be, for example, host physical addresses or trusted root partition guest physical addresses, as may depend on machine-specific DMA mechanisms. The table can also include additional fields as well, such as a field for measurements and at least one field for rule actions. As discussed above, rule actions can designate, for example, to lock a trusted root partition, leave or encapsulate / warp and encapsulate a packet.
[0092] An exemplary entry rules table can have multiple fields as well. For example, a key-matching field (target-matching) can be used for ingress rule matching, which can be one of the most complicated aspects of the system. In order not to have hardware that required a specific encapsulation format, a scheme could be used that is as generic as possible within what is reasonably achievable in hardware. FIG. 9 shows an example implementation of ingress matching key creation that can be used in accordance with a modality. The offload device can use multiple system-defined byte ranges and/or 904 byte range interleavers, which can be programmed by the trusted root partition at system startup, to interleave byte ranges from approaching 902 packets. These packets can be interspersed into a temporary byte 906 register, or other appropriate location. In at least one modality, four byte ranges from 0-128 bytes, with no more than 256 bytes from the beginning of the packet, may suffice, where all the byte ranges together total no more than 128 bytes. An additional 908 system wide bit mask (programmed by the trusted root partition) can then be applied to the byte register to determine which bytes are used to match the rules table. The final ingress matching key 910 can then be produced as a result, where the key can be used to look up the appropriate rule in the ingress rules table.
[0093] Other fields can be used with the ingress rules table as well. For example, a VM/VF ID field can be used, which can explicitly specify the VM/VF ID to forward to, where the rule action includes forwarding to a VM/VF. Other fields might include, for example, a measurement field and a rule action field, similar to the egress rules table discussed above. Measurements can be collected by hardware for later recovery by the trusted root partition. Examples of measurements that may be required by ingress/egress rule include the number of bytes acted on (left, forwarded, etc.) and the number of packets acted on (left, forwarded, etc.). Each metric field must be readable and clarifyable by the trusted root partition. The size of the fields can be at the discretion of the hardware vendor, for example, and can assume an interrupt-driven collection method from the trusted root partition.
[0094] In at least some modalities, there are at least two raw types of regulation or possible quality of service (QoS). A first type is referred to here as a type of “hard tops” regulation, in which each regulated entity is toped by a specific amount, regardless of the use of other regulated entities in the system. A second type is referred to here as a “surpassable tops” type, in which regulated entities are allowed to exceed their tops depending on whether there is excess capacity available in the system. In an exemplary modality, the ability to place hard tops may be required, such as at 50 Mb/s intervals (or 10-25 Mb/s intervals, in some arrangements) in SR-IOV Virtual Functions. In at least some modalities to use, in hardware, at least one regulation class per egress rule, so that different traffic can be regulated at different rates, and one QoS class per egress rule, so that different traffic can be prioritized. It may also be desirable in at least some embodiments to provide one or more classes of overrideable settings configurable by rule, so that unused capacity can be consumed if available and desired.
[0095] In at least some embodiments, it may be desirable to provide multiple checks on packages. For example, in some modalities, all egress packets must be checked for the correct L2 MAC address that was assigned to the VF. If the egress packet is an L3 IP, the source IP address must also be checked in at least some ways. Packets that do not have the correct L2 MAC and/or L3 IP address must be left in at least some modes. There may also be the ability to configure all L2 and/or L3 broadcast traffic to be locked to the trusted root partition, including IP broadcast and multicast, ARP, DHCP, etc. Furthermore, the trusted root partition in at least some embodiments will have the ability to inject ingress packets into virtual role packet queues. These packages can bypass the encapsulation/deformation system.
[0096] In at least some embodiments, the offload device hardware will support at least a standard set of offloads and hardware enhancements while performing encapsulation/deformation in virtual SR-IOV functions. These may include, for example, TCP segmentation offload (TSO) including the various checksum offloads, multi-queue capability, and interrupt coalescence. The suite can also include RDMA support (for example, RoCE or iWARP). Even if an L2 only RDMA Protocol is used, for example, the fact that the packet is encapsulated within an L3 packaging means that the application level protocol can be agnostic from the underlying physical network substrate.
[0097] The use of SR-IOV can neer a virtualization benefit where the underlying hardware is no longer abstracted. In order to preserve the same level of flexibility while providing advanced functionality to users, hardware vendors can provide approaches to dynamically inject driver code from the VMM into the guest VM. Such an approach might allow a single abstract driver in a guest VM to run on arbitrary hardware through a common interface, thus involving a hardware device completely emulated in software or one that is largely implemented in hardware.
[0098] In addition to those listed above, several other rules can be implemented as well. For example, for egress packets, there might be a list of allowable destination MAC addresses and destination IP subnets that form the ‘match’ part of each rule. A rule can have a destination MAC address and destination IP subnet, or the rule can have only one destination MAC address, in which case all IP addresses can be accepted. Each rule can have an opaque 'N' byte header, a source MAC address, and a target MAC address as part of the rule. When a rule is matched, the opaque 'N' byte header can be inserted before the original L2 header, and MAC addresses in the L2 header can be replaced with pre-specified values. New external L2 and L3 headers (eg MAC and IP) can be inserted in front of the opaque field with an external source IP address, external destination IP address, external destination MAC, and external source MAC from from the rules table. Optionally, the opaque header can include L2 and L3 headers, where the unloading device can fill in such fields, such as ID, length, checksum, and real-time flags. In some embodiments, the internal source and destination IP addresses are also replaceable, such as to allow for future virtualization of NAT, anycast, etc.
[0099] At least part of the processing and management can be performed by a software management interface operable to run on a trusted host platform, such as Xen Dom-0. Such an interface may include distributed services to load network specifications per tenant in real time, such as may include regulation, security groups, and partner components. The interface can instruct an offload component to perform per-tenant specifications (SR-IOV), for example. These commands can be processed in real time as specifications change. The interface can also perform extended rule management on an offload component basis in the event that the hardware or other offload component is unable to concurrently maintain all of the rules at any given time. These can include, for example, techniques such as loading hot rules, or a subset of frequently used rules, while processing a subset of less frequently used rules through software lock-in or other such process. The interface can differentiate between different types of traffic, such as traffic destined for the trusted host platform or a virtual tenant, and can deliver accordingly.
[00100] In at least some embodiments, packages that require special handling, such as Address Resolution Protocol (ARP) packages and multicast packages can also be managed by a software management component in Dom-0. Other advanced functionality such as DNS, a security interface, and a server web interface can also be handled by the software management interface. For a security interface, an instance can perform a secure connection before gaining network connectivity. The server web interface can be, for example, an interface to a metadata service or other such application.
[00101] Various embodiments can be described in view of the following clauses: 1. A FRAMEWORK for processing data packets in a multi-tenant environment, comprising: at least one processor; and memory including instructions that, when executed by the processor, allow the FRAMEWORK to: communicate with one or more distributed services to load one or more network specifications per tenant; instruct at least one offload device to perform tenant-loaded network specifications; manage a rule set for the at least one offload device when the at least one offload device is unable to concurrently store all of the rule set; and deliver data packets to an appropriate destination for each of a plurality of traffic types. 2. The FRAMEWORK of clause 1, wherein the FRAMEWORK provides a software management interface operable to run in a trusted host domain. 3. The FRAMEWORK of clause 2, where a software management interface is additionally operable to manage packages that require special processing. 4. The FRAMEWORK of clause 3, where packages that require special processing include MULTICAST packages, BROADCAST packages, and Address Resolution Protocol (ARP) packages. 5. The FRAMEWORK of clause 2, wherein the software management interface is additionally operable to manage functionality including at least one of domain name service (DNS), security interface, and web server interface. 6. The FRAMEWORK of clause 2, where the software management interface is operable to configure network statistics that need to be collected, and statistics to be kept, by the offload device. 7. The FRAMEWORK of clause 1, where network specifications per tenant include specifications for at least one of regulatory data packets, operating safety groups, and communicating between partner components. 8. The FRAMEWORK of clause 1, where network specifications per tenant are SR-IOV network specifications. 9. The FRAMEWORK of clause 1, where network specifications per tenant are processed in real time as specifications change. 10. The FRAMEWORK of clause 1, wherein managing a set of rules for the at least one unloading device includes loading a first subset of rules into a unloading device while processing a second subset of rules using the lashing of the software. 11. The FRAMEWORK of clause 10, where the first subset of rules is used more often than the second subset of rules. 12. The FRAMEWORK of clause 1, where the per-tenant specifications allow a hardware vendor to support multiple protocols without obtaining specific information about those multiple protocols. 13. The FRAMEWORK of clause 1, wherein the traffic types include at least one of traffic destined for a trusted host platform and traffic destined for a virtual tenant. 14. An unloading device, comprising: a processor; and memory storage instructions that, when executed by the processor, allow the offload device to: expose the offload device as a hardware device; perform at least the processing portion of a received user data packet for a physical function associated with the offload hardware device, the processing including at least removing an internal and external HEADER from the data packet, performing any modification of packet, and forwarding the user data packet to an internal virtual function, the internal virtual function operable to deliver the user data packet to a guest virtual machine. 15. The offload device of clause 14, wherein the processing includes removing at least one outer encapsulation HEADER from the user data packet. 16. The offload device of clause 14, wherein the offload device is a network interface card (NIC). 17. The offload device of clause 14, wherein the offload device is operable to support multiple protocols without obtaining specific information about those multiple protocols. 18. A method for processing data packets in a multi-tenant environment, comprising: communicating with one or more distributed services to load one or more network specifications per tenant; instructing at least one offload device to perform tenant-loaded network specifications; managing a rule set for the at least one offload device when the at least one offload device is unable to concurrently store all of the rule set; and delivering data packets to an appropriate destination for each of a plurality of traffic types. 19. The method of clause 18, further comprising: exposing a software management interface operable to run in a trusted host domain. 20. The method of clause 19, wherein the software management interface is additionally operable to manage functionality including at least one of domain name service (DNS), security interface, and web server interface. 21. The method of clause 19, wherein the discharge device operates in accordance with SR-IOV network specifications. 22. A computer-implemented method for processing data packets in an electronic environment, comprising: under the control of one or more computer systems configured with executable instructions, receiving a user data packet for a virtual function associated with a network virtual for a user; perform a rule table lookup for at least one rule for processing the user data packet; perform software-based processing of the user data packet in a trusted domain in response to determining a lock-in rule from the rules table; not perform any further processing of the user data packet in response to determining a release rule from the rules table; and performing at least a portion of the processing of the user data packet using a offload device in response to determining a routing rule from the rules table, the processing including at least adding an external HEADER to the data packet and send the user data packet out on a physical network, the external HEADER including at least one opaque field and including protocol-specific information. 23. The computer-implemented method of clause 22, further comprising: performing at least one generic check on the user data packet before performing the search. 24. The computer-implemented method of clause 23, wherein the at least one generic check includes at least one of level two (L2) or level three (L3) anti-masking, or trapping for at least one type of packet. 25. The computer-implemented method of clause 22, wherein the search is performed by the unloading device. 26. The computer-implemented method of clause 25, wherein the offload device provides a virtualized overlay network based on a single root I/O virtualization protocol (SR-IOV). 27. The computer-implemented method of clause 22, further comprising: performing at least one metric update on the user data packet before performing the search. 28. The computer-implemented method of clause 22, wherein the processing using a discharge device further includes at least one of the regulation the user data packet, or performance of a quality of service action. 29. The computer-implemented method of clause 22, wherein sending the user data packet out over a physical network is performed as part of a segmentation offload process. 30. The computer-implemented method of clause 22, wherein the processing using an offload device further includes updating the HEADER fields of the user data packet, the HEADER fields including at least one of an inner packet length, and an external, and an internal and an external checksum, a source and destination address, and a TIME-TO-LIVE (TTL) value. 31. The computer-implemented method of clause 22, wherein the processing using an offload device includes packet source checking performance on each egress packet based at least in part on a virtual machine source. 32. The computer-implemented method of clause 22, wherein the software-based processing includes processing by Dom-0 control software. 33. The computer-implemented method of clause 22, wherein the processing uses a generic format so that any appropriate protocol is capable of being supported by changing the parameters of a search key. 34. The computer-implemented method of clause 33, wherein the appropriate protocol is capable of being mapped to a stateless tunneling protocol. 35. A computer-implemented method for processing data packets in an electronic environment, comprising: under the control of one or more computer systems configured with executable instructions, receiving a user data packet for a physical function associated with a device discharge; constructing a search key for the user data packet using the download device; perform a rule table lookup for at least one rule for processing the user data packet using the lookup key; perform software-based processing of the user data packet in a trusted domain in response to determining a lock-in rule from the rules table; not perform any further processing of the user data packet in response to determining a release rule from the rules table; and perform at least a portion of the processing of the user data packet using the offload device in response to determining a routing rule from the rules table, the processing including at least removing an internal and external HEADER, performance from any package modification, and forwarding the user data package to an internal virtual function, the operable internal virtual function to deliver the user data package to a guest virtual machine. 36. The computer-implemented method of clause 35, wherein the processing using the offload device includes removing at least one outer encapsulation HEADER from the user data packet. 37. The computer-implemented method of clause 35, wherein the internal virtual function is identified by the routing rule. 38. The computer-implemented method of clause 35, wherein the processing using the offload device is operable to identify the user data packet as being encapsulated using a predefined protocol format at a predefined offset. 39. The computer-implemented method of clause 35, further comprising: processing the user data packet using software-based processing when the user data packet is not encapsulated. 40. The computer-implemented method of clause 35, wherein the offload device is a network interface card (NIC). 41. The computer-implemented method of clause 35, further comprising: determining a virtual machine corresponding to the user data packet using a fixed length field in the opaque bits at a predetermined offset in the user data packet. 42. The computer-implemented method of clause 35, wherein each physical function has a set of ingress rules, each rule consisting at least partially of a set of opaque bits capable of being matched with the opaque bits of encapsulated ingress packets. 43. The computer-implemented method of clause 35, wherein other traffic is traversed in response to determining a pass rule from the rule table. 44. The computer-implemented method of clause 35, in which encapsulated consumer traffic and control traffic is trapped independent of determining a passing rule from the rules table. 45. A system for processing data packets in an electronic environment, comprising: a processor; and a memory device including instructions which, when executed by the processor, cause the processor to: receive a user data packet for a virtual function associated with a virtual network for a user; perform a rule table lookup for at least one rule for processing the user data packet; perform software-based processing of the user data packet in a trusted domain in response to determining a lock-in rule from the rules table; do not perform further processing of the user data packet in response to determining a release rule from the rules table; and perform at least a portion of the user data packet processing using a offload device in response to determining a routing rule from the rules table, the processing including at least adding an external HEADER to the data packet and sending the user data packet out on a physical network, the external HEADER including at least one opaque field and including protocol-specific information. 46. The system of clause 46, further comprising: at least one discharge device operable to perform the search. 47. The system of clause 46, wherein the offload device provides a virtualized overlay network based on a single root I/O virtualization protocol (SR-IOV). 48. A system for processing data packets in an electronic environment, comprising: a processor; and a memory device including instructions which, when executed by the processor, cause the processor to: receive a user data packet for a physical function associated with a discharge device; build a search key for the user data packet using the download device; perform a rule table lookup for at least one rule for processing the user data packet using the lookup key; perform software-based processing of the user data packet in a trusted domain in response to determining a lock-in rule from the rules table; do not perform further processing of the user data packet in response to determining a release rule from the rules table; and perform at least a portion of the user data packet processing using the offload device in response to determining a forwarding rule from the rules table, the processing including at least removing an internal and external HEADER, the performance of any package modification, and forwarding the user data package to an internal virtual function, the internal virtual function operable to deliver the user data package to a guest virtual machine. 49. The system of clause 48, wherein the processing using a offload device includes removing at least one outer encapsulation HEADER from the user data packet. 50. The system of clause 48, wherein the offload device is a network interface card (NIC).
[00102] As discussed above, the various modalities can be implemented in a wide variety of operating environments, which in some cases may include one or more user computers, computing devices, or processing devices that can be used to operate any one of a number of applications. User or client devices can include any one of a number of general purpose personal computers, such as desktop or laptop computers running on a standard operating system, as well as mobile software running on cellular, wireless and portable devices. devices and capable of supporting a number of messaging and network protocols. Such a system may also include a number of workstations running on any one of a variety of commercially available operating systems and other known applications for purposes such as database development and management. These devices can also include other electronic devices, such as token terminals, thin-clients, gaming systems, and other devices capable of communicating over a network.
[00103] Various aspects can also be implemented as part of at least one service or web service, just as they can be part of a service-oriented architecture. Services such as services on the Web can communicate using any appropriate type of message, such as through the use of messages in Extensible Markup Language (XML) format and exchanged using an appropriate protocol, such as SOAP (derived from “Access Protocol of Single Object”). Processes provided or executed through such services may be written in any appropriate language, such as Web Service Description Language (WSDL). The use of a language such as WSDL allows functionality such as automated client-side code generation in various SOAP frameworks.
[00104] Most modalities use at least one network that would be familiar to those skilled in the art to support communications using any of a variety of commercially available protocols, such as TCP/IP, OSI, FTP, UPnP, NFS, CIFS, and AppleTalk. The network can be, for example, a local area network, a wide area network, a virtual private network, the internet, an intranet, an extranet, a public switched telephone network, an infrared network, a wireless network, and any of its combinations.
[00105] In embodiments utilizing a web server, the web server can run on any of a variety of server or middle tier applications, including HTTP servers, FTP servers, CGI servers, data servers, Java servers, and application servers of business. The server(s) may also be able to run programs or scripts in response to requests from a user's device, such as when running one or more web applications that can be deployed as a or more scripts or programs written in any programming language, such as Java®, C, C#, or C++, or any scripting language, such as Perl, Python, or TCL, as well as combinations thereof. The server(s) may also include database servers, including without limitation those commercially available from Oracle®, Microsoft®, Sybase®, and IBM®.
[00106] The environment may include a variety of data stores and other memory and storage media as discussed above. These can reside in a variety of locations, such as on a storage medium local to (and/or residing on) one or more of the computers or remote from any or all computers over the network. In a particular set of modalities, information may reside in a storage area network (“SAN”) familiar to those skilled in the art. Similarly, any files necessary to perform the functions assigned to computers, servers, or other network devices may be stored locally and/or remotely, as appropriate. Where a system includes computerized devices, each such device may include hardware elements that may be electrically coupled via a bus, the elements including, for example, at least one central processing unit (CPU), at least one input device (per for example, a mouse, keyboard, controller, touch screen, or numeric keypad), and at least one output device (for example, a display device, printer, or speaker). Such a system may also include one or more storage devices, such as disk drives, optical storage devices, and solid-state storage devices, such as random access memory ("RAM") or read-only memory (" ROM”), as well as removable media devices, memory cards, flash cards, etc.
[00107] Such devices may also include a computer-readable storage media reader, a communications device (eg, a modem, a network card (wired or wireless), an infrared communication device, etc.), and working memory as described above. The computer readable storage media reader may be connected to, or configured to receive, a computer readable storage medium representing remote, local, fixed, and/or removable storage devices, as well as storage media to contain, store, transmit and retrieve temporarily and/or more permanently computer-readable information. The system and various devices will also typically include a number of software applications, modules, services, or other elements located within at least one working memory device, including an operating system and application programs, such as a client application. or web browser. It should be appreciated that alternative modalities have numerous variations from the one described above. For example, custom hardware may also be used and/or particular elements may be implemented in hardware, software (including portable software such as applets), or both. Furthermore, connection to other computing devices such as network input/output devices can be employed.
[00108] Storage media and computer readable media to contain code, or portions of code, may include any appropriate media known or used in the art, including storage media and communication media such as, but not limited to, volatile media and non-volatile, removable and non-removable implemented in any method or technology for storing and/or transmitting information, such as computer readable instructions, data structures, program modules, or other data, including RAM, ROM, EEPROM, memory, flash or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that may be used to store the information that is desired and accessible by a system device. Based on the disclosure and teachings provided here, a person of ordinary skill in the art will appreciate other ways and/or methods for implementing the various modalities.
[00109] The specification and figures should correspondingly be considered in an illustrative rather than a restrictive sense. It will, however, be evident that various modifications and alterations can be made to this without departing from the broader spirit and scope of the invention as set out in the claims.

权利要求:
Claims (15)
[0001]
1. Discharge device (506), characterized in that it comprises: a processor (512); and memory (510) storing instructions which, when executed by the processor, allow the offload device to: operate in accordance with SR-IOV network specifications; expose the offload device as a hardware device; constructing a search key for a user data packet received for a physical function associated with the download device using the download device; performing a search (706) on a rules table for at least one rule for processing the user data packet using the search key; perform software-based processing (712) of the user data packet in a trusted domain in response to determining a lock-in rule from the rules table; not perform any further processing (714) of the user data packet in response to determining a release rule from the rules table; and performing at least a processing portion of the user data packet in response to the forwarding rule determination from the rules table, the processing including at least removing an external HEADER from the data packet, performing any packet modification, and forward the user data package to an internal virtual function of the offload device, the internal virtual function operable to deliver the user data package to a guest virtual machine.
[0002]
2. Unloading device according to claim 1, characterized in that the processing includes the removal of at least one external encapsulation HEADER from the user data packet.
[0003]
3. Discharge device according to claim 1, characterized in that the discharge device is a network interface card (NIC).
[0004]
4. Discharge device according to claim 1, characterized in that the discharge device is operable to support multiple protocols without obtaining specific information about those multiple protocols.
[0005]
5. Computer-implemented method for processing data packets in an electronic environment, characterized by the fact that it comprises: under the control of one or more computer systems configured with executable instructions, receiving a user data packet for a physical function associated with a discharge device (506) that operates in accordance with SR-IOV network specifications; constructing a search key for the user data packet using the download device; performing a search (706) on a rules table for at least one rule for processing the user data packet using the search key; perform software-based processing (712) of the user data packet in a trusted domain in response to determining a lock-in rule from the rules table; not perform any further processing (714) of the user data packet in response to determining a release rule from the rules table; and perform at least a portion of the user data packet processing using the offload device in response to determining a routing rule from the rules table, the processing including at least removing an external HEADER, performing any packet modification , and forwarding (716) the user data packet to an internal virtual function, the internal virtual function operable to deliver the user data packet to a guest virtual machine.
[0006]
6. Computer-implemented method according to claim 5, characterized in that the processing using the offloading device includes the removal of at least one external encapsulation HEADER from the user data packet.
[0007]
7. Computer-implemented method, according to claim 5, characterized by the fact that the internal virtual function is identified by the forwarding rule.
[0008]
8. The computer-implemented method of claim 5, characterized in that the processing using the offload device is operable to identify the user data packet as being encapsulated using a predefined protocol format at a predefined offset .
[0009]
9. Computer-implemented method according to claim 5, characterized in that the processing using a discharge device additionally includes updating HEADER fields of the user data packet, the HEADER fields including at least one of one inner and one outer packet length, and one inner and one outer checksum, one source and one destination address, and a TIME-TO-LIVE (TTL) value.
[0010]
10. Computer-implemented method according to claim 5, characterized in that it further comprises: processing the user data packet using software-based processing when the user data packet is not encapsulated.
[0011]
11. Computer-implemented method according to claim 5, characterized in that the offloading device includes a network interface card (NIC).
[0012]
12. The computer-implemented method of claim 5, further comprising: determining a virtual machine corresponding to the user data packet using a fixed-length field in the opaque bits at a predetermined offset in the data packet of user.
[0013]
13. Computer-implemented method according to claim 5, characterized in that each physical function has a set of ingress rules, each rule consisting of at least partially a set of opaque bits capable of being matched with bits opaque of encapsulated ingress packets.
[0014]
14. Computer-implemented method according to claim 5, characterized in that other traffic is traversed without processing in response to the determination of a passing rule from the rules table.
[0015]
15. Computer-implemented method, according to claim 5, characterized by the fact that the encapsulated traffic of consumer and control traffic is trapped independent of the determination of a passing rule from the rules table.

类似技术:

---

公开号 | 公开日 | 专利标题

---

US11099885B2|2021-08-24|Frameworks and interfaces for offload device-based packet processing

---

US9602636B1|2017-03-21|Stateless packet segmentation and processing

---

US9042403B1|2015-05-26|Offload device for stateless packet processing

---

BR112013024883B1|2021-06-29|FRAMEWORKS AND INTERFACES FOR UNLOADING DEVICE-BASED PACKAGE PROCESSING

---

US8155146B1|2012-04-10|Stateless packet segmentation and processing

---

US9712538B1|2017-07-18|Secure packet management for bare metal access

---

US9385912B1|2016-07-05|Framework for stateless packet tunneling

---

US20200344088A1|2020-10-29|Network interoperability support for non-virtualized entities

同族专利:

---

公开号 | 公开日

---

AU2012236513A1|2013-10-31|

---

CA2831705A1|2012-10-04|

---

CA2951949C|2019-01-15|

---

CA2951952C|2019-01-15|

---

AU2012236513A8|2013-12-05|

---

CA2951970A1|2012-10-04|

---

BR112013024883A2|2016-12-20|

---

JP2018011331A|2018-01-18|

---

JP5869099B2|2016-02-24|

---

SG194017A1|2013-11-29|

---

JP6487979B2|2019-03-20|

---

AU2012236513B2|2015-02-05|

---

JP2014512760A|2014-05-22|

---

CN104054067B|2017-09-08|

---

JP2017126998A|2017-07-20|

---

CA2831705C|2017-10-03|

---

CA2951949A1|2012-10-04|

---

CN107450966B|2021-08-06|

---

JP2016028479A|2016-02-25|

---

CA2951970C|2018-02-13|

---

JP6207559B2|2017-10-04|

---

WO2012135442A1|2012-10-04|

---

JP6360576B2|2018-07-18|

---

CN104054067A|2014-09-17|

---

CN107450966A|2017-12-08|

---

EP2691865A1|2014-02-05|

---

CA2951952A1|2012-10-04|

---

EP2691865A4|2016-05-25|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

US5802258A|1996-05-03|1998-09-01|International Business Machines Corporation|Loosely coupled system environment designed to handle a non-disruptive host connection switch after detection of an error condition or during a host outage or failure|

---

US7269663B2|2001-09-28|2007-09-11|Intel Corporation|Tagging packets with a lookup key to facilitate usage of a unified packet forwarding cache|

---

JP2003303053A|2002-04-11|2003-10-24|Nec Corp|Disk array apparatus and data processing method using same|

---

US7496689B2|2002-04-22|2009-02-24|Alacritech, Inc.|TCP/IP offload device|

---

US7701963B2|2002-10-15|2010-04-20|Qualcomm Incorporated|Method and apparatus for the use of micro-tunnels in a communications system|

---

US7620070B1|2003-06-24|2009-11-17|Nvidia Corporation|Packet processing with re-insertion into network interface circuitry|

---

EP1515511B1|2003-09-10|2011-10-12|Microsoft Corporation|Multiple offload of network state objects with support for failover events|

---

US7961683B2|2004-09-30|2011-06-14|Alcatel-Lucent Usa Inc.|Active session mobility solution for point-to-point protocol|

---

US7957379B2|2004-10-19|2011-06-07|Nvidia Corporation|System and method for processing RX packets in high speed network applications using an RX FIFO buffer|

---

US7783880B2|2004-11-12|2010-08-24|Microsoft Corporation|Method and apparatus for secure internet protocol offloading with integrated host protocol stack management|

---

US7797460B2|2005-03-17|2010-09-14|Microsoft Corporation|Enhanced network system through the combination of network objects|

---

JP4616732B2|2005-09-02|2011-01-19|株式会社日立製作所|Packet transfer device|

---

US8069153B2|2005-12-02|2011-11-29|Salesforce.Com, Inc.|Systems and methods for securing customer data in a multi-tenant environment|

---

JP5022691B2|2006-12-12|2012-09-12|キヤノン株式会社|COMMUNICATION DEVICE, ITS CONTROL METHOD, AND PROGRAM|

---

CN101207604B|2006-12-20|2012-03-28|联想有限公司|Virtual machine system and communication processing method thereof|

---

US7715321B2|2007-01-30|2010-05-11|International Business Machines Corporation|Network interface card transmission control protocol acceleration offload failure detection and recovery mechanism|

---

US8412809B2|2007-10-24|2013-04-02|International Business Machines Corporation|Method, apparatus and computer program product implementing multi-tenancy for network monitoring tools using virtualization technology|

---

CN101179509A|2007-12-10|2008-05-14|中兴通讯股份有限公司|Point-to-point protocol based MAC address and access circuit binding method and apparatus|

---

US20090199290A1|2008-02-01|2009-08-06|Secure Computing Corporation|Virtual private network system and method|

---

JP2009278261A|2008-05-13|2009-11-26|Toshiba Corp|Information processing device and communication control method|

---

CN101753546B|2008-12-16|2012-10-31|中国移动通信集团公司|Data packet transmission method and device|

---

CN101459618B|2009-01-06|2011-01-19|北京航空航天大学|Data packet forwarding method and device for virtual machine network|

---

US8589919B2|2009-04-28|2013-11-19|Cisco Technology, Inc.|Traffic forwarding for virtual machines|

---

EP2509000A4|2009-12-04|2017-09-20|Nec Corporation|Server and flow control program|

---

US10353722B2|2010-07-21|2019-07-16|Nec Corporation|System and method of offloading cryptography processing from a virtual machine to a management module|

---

CN101969475A|2010-11-15|2011-02-09|张军|Business data controllable distribution and fusion application system based on cloud computing|US8774213B2|2011-03-30|2014-07-08|Amazon Technologies, Inc.|Frameworks and interfaces for offload device-based packet processing|

---

US9935841B2|2013-01-28|2018-04-03|Intel Corporation|Traffic forwarding for processing in network environment|

---

US9384025B2|2013-01-28|2016-07-05|Intel Corporation|Traffic and/or workload processing|

---

US9426154B2|2013-03-14|2016-08-23|Amazon Technologies, Inc.|Providing devices as a service|

---

US9747249B2|2014-12-29|2017-08-29|Nicira, Inc.|Methods and systems to achieve multi-tenancy in RDMA over converged Ethernet|

---

CN106528362B|2015-09-10|2019-03-19|阿里巴巴集团控股有限公司|A kind of flow partition method and device|

---

US10419772B2|2015-10-28|2019-09-17|Qualcomm Incorporated|Parallel arithmetic coding techniques|

---

CN107291525B|2016-04-01|2021-06-01|华为技术有限公司|Method, host machine and system for deploying virtual machine|

---

CN106411872B|2016-09-21|2019-09-17|杭州迪普科技股份有限公司|A kind of method and apparatus of the message compression based on Packet Classification|

---

US10785020B2|2018-01-19|2020-09-22|Microsoft Technology Licensing, Llc|Hardware offload for QUIC connections|

---

CN112717380A|2019-03-19|2021-04-30|Oppo广东移动通信有限公司|Network detection method and related device|

---

KR102217114B1|2020-07-24|2021-02-18|넷록스 주식회사|Method for controlling of accelerating edge platform network and electronic device using the same|

---

US20220038371A1|2020-07-30|2022-02-03|S.C Correct Networks S.R.L.|Method for transferring information across a data center network|

---

CN112600826B|2020-12-10|2022-02-22|中国科学院深圳先进技术研究院|Virtualization security gateway system|

法律状态:
2018-12-18| B06F| Objections, documents and/or translations needed after an examination request according [chapter 6.6 patent gazette]|

---

2019-10-22| B06U| Preliminary requirement: requests with searches performed by other patent offices: procedure suspended [chapter 6.21 patent gazette]|

---

2021-04-20| B09A| Decision: intention to grant [chapter 9.1 patent gazette]|

---

2021-06-29| B16A| Patent or certificate of addition of invention granted [chapter 16.1 patent gazette]|Free format text: PRAZO DE VALIDADE: 20 (VINTE) ANOS CONTADOS A PARTIR DE 29/03/2012, OBSERVADAS AS CONDICOES LEGAIS. |

---

2021-11-03| B16C| Correction of notification of the grant [chapter 16.3 patent gazette]|Free format text: REFERENTE AO DESPACHO 16.1 PUBLICADO NA RPI 2634, QUANTO AO DEPOSITANTE |

优先权:

---

申请号 | 申请日 | 专利标题

---

US13/076,339|2011-03-30|

---

US13/076,347|2011-03-30|

---

US13/076,339|US8462780B2|2011-03-30|2011-03-30|Offload device-based stateless packet processing|

---

US13/076,347|US8774213B2|2011-03-30|2011-03-30|Frameworks and interfaces for offload device-based packet processing|

---

PCT/US2012/031121|WO2012135442A1|2011-03-30|2012-03-29|Frameworks and interfaces for offload device-based packet processing|

---