• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    Method of routing data in a network, this data coming from a first node and being destined for a second node, this method comprising the following steps: pre-registration in each of the switches of a list of directly interfaced switches with one or more nodes; reception of the data by a first switch and determination of its destination; when the first switch is directly interfaced with the second node via an output port - selection and routing of the data by said output port; otherwise - identification of the switch to which the second node is directly interfaced; routing the data to the identified switch via an output port of the first switch, so that the data can be routed to said identified switch.
    公开号:FR3022423A1
    申请号:FR1455509
    申请日:2014-06-16
    公开日:2015-12-18
    发明作者:Pierre Vigneras；Jean-Noel Quintin
    申请人:Bull SA；
    IPC主号:
    专利说明:

    [0001] The invention relates to the field of routing through a network. More particularly, the invention relates to the management of tables allowing the routing of information flows through a network of a supercomputer. Appeared in the sixties, the progressive evolution of HPC supercomputers (acronym for "High Performance 10 Computing"), allows years and years to provide computational powers and data processing capabilities ever more important. These supercomputers are commonly used to model and simulate complex phenomena that can not be achieved by experimental approach. Supercomputers are therefore used for a wide variety of applications such as weather simulations, financial modeling, probabilistic analyzes, aeronautical, medical, physical, molecular or nuclear simulations. A supercomputer is typically realized by a network of 20 nodes interconnected by switches ("switches" in English). A node may, by way of example, be a computer, a server or any other computer terminal comprising a plurality of microprocessors, as well as means for transmitting / receiving data. Advantageously, these nodes make it possible to receive or transmit data (eg messages, packets, datagram) via one or more network devices, for example using a network card. Switches have the function of routing the data to and from the nodes to which they are connected. Advantageously, the nodes and switches constitute a computer network or a graph according to a predetermined topology. Currently, in the context of supercomputers, there are few networks using adaptive routing. Commonly, to achieve adaptive routing, each switch keeps in a routing table information about the different possibilities for any packet to reach any possible destination. B021 B005 EN Version: TQD The creation of a routing table is the result of a software program in charge of calculating all the possible paths of a packet to a destination. As an example, the Dijkstra algorithm is commonly implemented to compute the set of shortest paths in a graph. Generally, in order to avoid recalculating the paths each time a calculation algorithm is modified, all the possible paths for all the nodes of the graph are prerecorded in the memory space of the switches. This results in very large routing tables for graphs with limited complexity. Supercomputer graph or network topologies are highly complex due to the large number of interconnected nodes and switches. The number of possible recipients and path possibilities being limited by the available memory space in each switch, such a solution is limited because it does not cover all routing possibilities. Other known proposals are to consider hierarchical routing solutions. As part of these solutions, it is no longer a single destination, but a subset of a plurality of destinations. It is thus possible to choose a group comprising a limited number of destinations. However, such proposals are also limited. It is indeed impossible to determine all the possibilities of paths to a destination, because of the limitation in a switch of the memory space necessary for storing the calculations, as well as the calculation time required to determine this information. A first objective is to remedy the aforementioned drawbacks. A second objective is to propose routing tables making it possible to cover all the possible routes towards a given destination. A third objective is to propose an adaptive routing solution, making it possible to overcome computational time constraints and limitations of calculation memory space in a switch. B021 B005 GB Version: TQD A fourth objective is to limit the amount of memory occupied by the routing tables in the switches of a network or a graph. For this purpose, it is proposed, according to a first aspect, a method of routing data in a network comprising a plurality of switches and a plurality of nodes, this data coming from a first node and being destined for a second node. , this method comprising the following steps: pre-recording in each of the switches of the network, a list of switches directly interfaced with one or more nodes; receiving the data by a first switch via an input port; determining the destination of said data by the first switch; when the first switch is directly interfaced with the second node via a determined output port selection and routing of the data by said determined output port; otherwise - identification of the switch to which the second node is directly interfaced by comparison between the destination of the data and the prerecorded switch list; routing the data to the identified switch via an output port of the first switch, so that the data can be routed to said identified switch. Advantageously, in this method, the selection of the output port to the identified switch is performed by a comparison step between the identified switch and a routing table of the first switch, each entry of this table comprising - information enabling identify a switch interfaced with at least one node; B021 B005 EN Version: TQD - a set of aggregated information describing at least one route through a specified port to that switch interfaced with at least one node. Advantageously, in this method, for a plurality of possible routes to a switch interfaced with at least one node, the selection of the output port is performed according to the aggregated information, this information including the number of jumps, the congestion or the weighting of the links or nodes on the road to reach the switch via a specific port.
    [0002] Advantageously, in this method, the network is a supercomputer topology. It is proposed, according to a second aspect, a computer program product implemented on a memory medium, capable of being implemented within a computer processing unit and including instructions for the implementation of the summary method. above. It is proposed, in a third aspect, a switch in a network comprising a plurality of switches and a plurality of nodes, this switch comprising a plurality of input and output ports, able to receive and route data from a first node to a second node; a prerecorded list of switches directly interfaced with one or more nodes; means capable of determining the destination of said datum; means capable of selecting and then routing the data by a determined output port, when the switch is directly interfaced with the second node via said determined output port; means capable of identifying a second switch to which the second node is directly interfaced by comparison between the destination of the data and the prerecorded list of switches; means adapted to select and then route the data to the second identified switch via an output port, so that the data can be routed to said identified second switch. Advantageously, in this switch, the selection of the output port to the identified switch is performed by means able to compare the identified second switch with the entries of a prerecorded routing table, each entry of this table comprising - information enabling to identify a switch interfaced with at least one node; a set of aggregated information describing at least one route by a specific port to said switch interfaced with at least one node. Advantageously, this switch comprises means capable of selecting for a plurality of possible routes to a switch interfaced with at least one node, an output port according to the aggregated information, this information comprising for a given port the number of jumps, the congestion or weighting the links or nodes on the route to reach the switch interfaced with at least one node.
    [0003] It is proposed, in a fourth aspect, a network of a supercomputer, comprising a plurality of switches and a plurality of nodes, the switches being made in the manner summarized above. Other objects and advantages of the invention will emerge in the light of the description of an embodiment, given hereinafter with reference to the accompanying drawings, in which FIG. 1 is an example of a simplified topology of a network or a graph, comprising a plurality of nodes and a plurality of switches. According to various embodiments, a network or a graph comprising a plurality of nodes and switches arranged according to a predetermined topology is considered. Advantageously, this network or graph relates to the topology of a supercomputer, or to any other network or graph having a similar topology. A node may, by way of example, be a computer terminal, comprising means for transmitting and receiving data, such as one or more network cards, as well as data processing means, for example: example a plurality of microprocessors. Advantageously, the nodes allow: to execute instructions received via an interface, for example to conduct a succession of calculations. We then speak of knots of computation; - manage the input / output of data storage systems. These are known as service, administration or storage nodes. The switches comprise a plurality of input and output ports, and can be interfaced directly with other switches. By "directly interfaced" is meant here the fact that two entities, here switches, each comprising at least one interface (eg input / output port), are connected by a link arranged between each of these interfaces, by example, a physical or virtual link; - Interconnect different nodes of the network or the graph, and ensure the routing / switching of data from or to the different nodes. These switches are referred to below as leaf switches. FIG. 1 illustrates a very simplified example of a network or graph topology, comprising a plurality of nodes 0, 1, 2, 3, 4, 5, 6, 7 and a plurality of switches a, b, c, d. , e, f. In this example: - the nodes 0 and 1 are interfaced with the switch a - - the nodes 2 and 3 are interfaced with the switch b - the nodes 4 and 5 are interfaced with the switch c - the nodes 6 and 7 are interfaced with the switch d - the switch e is interfaced with the switches a, b, c, d; the switch f is interfaced with the switches a, b, c, d; - only the switches a, b, c, d are leaf switches. A network or graph is then considered, for example modeling the topology of a supercomputer, comprising a first node connected to a first switch and a second node connected to a second switch. B021 B005 EN Version: TQD Typically, when a first switch receives an input port with a datagram (or other data) from a first node to a second node, the first switch checks its table Routing. If the second node is interfaced directly via an output port with the first switch, for example via a wired link, the first switch transmits the packet. In the opposite case, said second node is interfaced with another switch, that is to say a second switch. The first switch then seeks to identify a route to the second destination node. The switch's routing table then allows adaptive routing by selecting a suitable output port. Here, adaptive routing means the choice of a path among a plurality in order to reach the destination which takes into account the state of the network, for example its congestion. To do this, the switch routing table is commonly formed of a plurality of entries, each entry may include one or more routes to a destination node. An input therefore comprises information relating to one or more output ports that can be selected. The choice of the output port, and therefore of the routing, is commonly determined according to various characteristics, as examples: congestion of the nodes or links (eg cables) in the network, number of jumps necessary to reach the node destination ("hop-count" in English), weighting allocated to different nodes or intermediate switches. Advantageously, all of these characteristics are stored in the prerecorded routing table for each of the switches. For example, consider for Figure 1 the switch a.
    [0004] In this example, this switch comprises four ports 10, 11, 12, 13 that can be used as input or output ports, depending on the direction of data flow in the network. In the current state of the art, a very simplified representation of the routing table of this switch is in the following form: B021 B005 EN Version: TQD Destination node Output port number (s) (s) ( s) associated 0 11 1 10 2 12; 13 3 12; 13 4 12; 13 12; 13 6 12; 13 7 12; 13 Each port number in this example is also associated with information (non-proprietary). shown) relating to the route to reach the destination, for example the number of jumps required, or any other feature previously mentioned. Thus, when a switch receives a datagram for a given destination, it makes use of its routing table a correspondence between the destination and a port number, this port number can be chosen according to the characteristics of the route to the destination node.
    [0005] The structure of this type of table nevertheless has many disadvantages, in particular in terms of memory space occupied in each switch and route possibilities. For example, considering that this routing table has eight inputs and two recordable ports at most per input, it is found that the switch e of Figure 1, with such a table structure is unable to cover all of possible combinations of routes to a destination. Indeed, even if the switch d proposes a shorter path in terms of topology towards a destination, routes passing through the switches a, b, c are also valid. Therefore, in order to cover all the possibilities of routing to a destination node, according to various embodiments, each entry of the routing table of a switch is made so as to be able to reach a leaf switch attached to a switch. destination group, i.e., a plurality of nodes. To do this, for each entry in the table, all the information allowing to reach the same switch attached to a group of destinations is aggregated. For example, one aggregates for the same entry, the numbers (or any identification information) of ports identified as available / usable for routing to a leaf switch directly interfaced to an identified destination, as well as the characteristics of the paths to reach. this leaf switch. A very simplified representation of such a routing table, for the switch a of Figure 1 in is given below: Identification of the number (s) of switch port (s) output (s) associated (s) of destination with empty fields b 12; 13c12; 13 of 12; Advantageously, the identification of a leaf switch to which to route a datagram is established by each switch. For example, each switch determines information about the destination of the received datagram, such as the destination address of the datagram (and potentially the address of the leaf switch in addition), and based on this information identifies the destination leaf switch. This translation can be done through a table or a calculation on the destination address. For example, the switch identifies the address of the leaf switch to route the datagram, via a comparison step between the destination of the datagram and a list of known leaf switches, which are for example prerecorded or pre-identified in a data table. . In another example, assuming that each leaf switch is numbered and interfaced directly with n nodes, a division by n of the identifier (eg, address) of a destination node identifies the leaf switch number. For example, for 2 leaf switches "0" and "1" each interfaced with 8 nodes, having for respective identifiers "0-7" and "8-15", B021 B005 FR Version: TQD the identification of the switch number is determined by dividing by 8 the identifier of a node and then taking the integer part of this result. In addition, each entry of the table includes for each port number, aggregate characteristics relating to the path to reach the leaf switch via said port number. For example: the number of hops to a destination leaf switch, the availability of physical links on the road or the congestion of intermediate nodes. Advantageously, these characteristics enable the switch to perform adaptive routing by selecting an output port proposing a route with optimal network characteristics. These characteristics may for example be chosen during the network deployment, determined during a first phase of exploration using an algorithm or evaluated periodically in the network, for example via measurements or dynamic algorithms. routing providing feedback on network characteristics. Thus, each entry in the routing table no longer describes one or more routes to a destination node, but one or more routes to a destination leaf switch, which switch serves a plurality of destinations. Advantageously, only the leaf switch of a destination group knows the characteristics (ex: number of the output port to use) allowing the final routing of a datagram to a specific node to which it is directly interfaced. According to one embodiment, the characteristics relating to this final routing are for example initially known and prerecorded in a table specific to each switch. A simple correspondence between the table and the destination of the datagram, allows the switch to determine the output port to which the datagram is to be transferred. A simplified representation of such a switch-specific table of Figure 1 in is given below: Destination Node Associated Output Port (s) Number (s) 0 11 1 10 B021 In another embodiment, each switch is configured to be able to map an output port to a node to which it is attached without necessarily having such a table. For example, the switch is able to query via a request a node on its identity, such as on its identification number, via the physical link directly interconnecting an output port and said node, and subsequently establish a correspondence between the number port and the identified node. Thus, when a first switch receives on an input port a datagram to be routed to a destination node, it performs the following operations: it identifies by correspondence, for example via a step of comparison between one of its tables and the field specifying the datagram destination, a destination node to which it is attached, or else a second switch to which the destination node is attached; if the switch is directly interfaced via an output port to the destination node, it routes the datagram through that port to the destination node; Otherwise, the switch identifies a second leaf switch connected to the destination node. It then establishes a correspondence between an entry of its routing table relating to the second leaf switch and at least one output port relating to this input; If the input concerned comprises a plurality of output ports, the switch then determines the best output port according to the characteristics of each port, for example by selecting the port that proposes the minimum of hops to the second leaf switch. ; Finally, the switch transmits the datagram to the best identified output port. Moreover, as can be seen in the example of the previous routing table, several entries relating to different target switches, may refer to the same numbers of output ports. The switch therefore retains the aggregated information for each of its ports, and when it receives a datagram to retransmit, is able to retransmit it on each of its ports. Advantageously, this characteristic makes it possible, compared to the state of the art, to no longer limit a destination to a fixed number of ports: the limit number of possible routes is here limited only by the maximum number of output ports of each switch. . In addition, if it is desired to further reduce the memory space occupied by the routing tables, it is also possible for a network or graph comprising a large number of leaf switches to aggregate a set of information. relating to routes to a leaf switch group. A first example relating to the previously described embodiments is given below. We consider here the hardware realization of a switch in a network with adaptive routing: the switch chooses for a given destination an output port, among the entries of its routing table. This switch includes 48 input / output ports, and for each destination 3 adaptive routes. In the state of the art, each route is defined by 6 bits including the encoding of the output port number. Therefore, each destination requires 3 * 6 = 18 bits in the switch routing table. Now, for the previously described embodiments, choose a 48-bit array for each input, i.e., for each leaf switch connecting a destination group. As soon as a group contains more than three destinations, the amount of memory occupied in the routing table is diminished: the memory gain for each entry is indeed here 3 * 6 * 3/48, or 9/8 per 30 entrance. Moreover the number of routing possibilities which was 3 * 3 adaptive routes in the state of the art passes to a maximum number limited by the number of ports of the switch, so here 48. A so-called thick tree topology ( "Fat-trees" in English) commonly used in supercomputer networks, refers to the generalized PGFT (Parallel Port Fat-Trees) topology. In these topologies, the number of B021 B005 FR Version: TQD computation nodes connected to the same switch is half the number of links, ie 24 computation nodes in this example. For the calculation nodes, therefore, we obtain a global memory gain for the routing table which is here 24 * 6 * 3/24, ie a memory gain of factor 18.
    [0006] In addition, the higher the number of nodes present on the computer, the more the memory gain is too. Advantageously, the routing possibilities are also improved: if one or more physical connection links or switches fail, it will indeed be possible for the switch directly connected to the fault to continue to route the messages without intervention or external communication as that it remains a usable link to reach the destination in its table. In a second example, the software implementation of switches performing a data routing or a calculation of the shortest paths in a graph simulating a constituent network of a supercomputer is considered. Typically, in the algorithms for calculating the shortest paths, such as the Dijkstra algorithm, each of the switches of a graph is considered as a vertex of the graph, and stores in its "routing table" (generally called the path table the shorter in the context of an algorithm for calculating the shortest paths) "for each destination the neighbor proposing the shortest path. Thus, from each switch of the graph, the shortest path is known by passing from neighbors to neighbors, and from the set of several neighbors can lead to the destination with the same number of hops. In accordance with the previously described embodiments, an algorithm for calculating the shortest paths is implemented in each switch. Advantageously, in this algorithm, instead of storing in the routing table of each switch a link to the shortest paths, for each output port the characteristics of the best path passing through this port are memorized to a leaf switch connecting a group of destination nodes. By way of example, the number of hops between the current switch and the leaf switch is stored as a characteristic. B021 B005 EN Version: TQD Advantageously, as a leaf switch potentially connects a large number of destination nodes, such an algorithm does not require additional memory for the routing table. Moreover, as in the first example, if a communication link is cut, the switch remains able to select one of the other links providing such a short path. Advantageously, such an algorithm implementation in a switch comprises a very limited memory complexity. Indeed, by noting g_ the number of leaf switches connecting groups of destination nodes, has the maximum number of edges connected to the same node and n the total number of nodes in the graph, we then obtain a complexity in memory of 0 (in * a). The Relational Reinforcement Learning (RRL) algorithm of Ramalingam & Reps, which is a variant of the Dijkstra algorithm, typically has a memory complexity of 0 (nA2 * a). Considering the embodiments described above, this complexity in memory is reduced to 0 (c * n * a) with c the number of calculation nodes. Thus, if we consider as an example that each leaf switch comprises 24 computation nodes, we obtain a memory gain of 24. Moreover, the computation time complexity for each input is divided by the number of destinations within the same group, that is to say the number of nodes attached to the same leaf switch.
    [0007] In this example, the previously proposed embodiments are compared with the RRL algorithm of Ramalingam & Reps, a graph being modeled with the following parameters: Total number of switches 9792 Number of calculation nodes 55296 Number of leaf switches 2304 Number of switches edges in the graph 200448 Number of edges between switches 145152 The results obtained for this graph modeling, including the construction of the correspondence tables for each of the switches, make it possible to obtain a memory occupancy of approximately 10 Gb with the proposed modes of realization, against approximately 1To for the RRL algorithm of Ramalingam & Reps. By observing the values of the preceding parameters proposed, it is understood that these results come from the fact that it is more advantageous to work on the leaf switches rather than on the computation nodes, the latter being much more numerous in the graph. Advantageously, considering a routing table comprising entries relating to routes between a switch and leaf switches, vis-à-vis a switch-type routing to nodes or routing type node nodes, includes many advantages: - a simplification of the topology from the point of view of each switch, it is not considered more individually each destination, but groups of destinations; - we get rid of the problems of congestion of knots. Indeed, generally a routing procedure proves ineffective to solve the problems of congestions of nodes. When a node fails to process all of the data it receives, balancing the routes does not solve anything: routing is the best way to reduce the number of adaptive routes to the congesting node; - The nodes are typically exponentially more numerous than the switches and therefore statistically more likely to fail. Considering the proposed modes of realization, the disappearance or the reappearance of nodes in a network or a graph, does not imply changes of perception of topologies from the point of view of the switches; the number of switches being typically in logarithm of the number of nodes, and the number of leaf switches in logarithm of the number of switches, the complexity in memory as well as the computational complexity are considerably reduced. B021 B005 FR Version: TQD
    权利要求:
    Claims (9)
    [0001]
    REVENDICATIONS1. A method of routing data in a network comprising a plurality of switches (af) and a plurality of nodes (0-7), this data coming from a first node and being destined for a second node, this method comprising the following steps: pre-registration in each of the switches of the network, a list of switches directly interfaced with one or more nodes; Receiving the data by a first switch via an input port; determining the destination of said data by the first switch; when the first switch is directly interfaced with the second node via a determined output port selection and routing of the data by said determined output port; otherwise, identification of the switch to which the second node is directly interfaced by comparison between the destination of the data item and the prerecorded list of switches; routing the data to the identified switch via an output port of the first switch, so that the data can be routed to said identified switch.
    [0002]
    The method of claim 1, wherein the selection of the output port to the identified switch is performed by a step of comparing the identified switch with a routing table of the first switch, each entry of that table comprising a information for identifying a switch interfaced with at least one node; a set of aggregated information describing at least one route by a specific port to said interfaced switch with at least one node. B021 B005 FR Version: TQD
    [0003]
    3. Method according to claim 2, wherein for a plurality of possible routes to a switch interfaced with at least one node, the selection of the output port is performed according to the aggregated information, this information including the number of jumps, congestion or the weighting of links or nodes on the route to reach the switch via a specific port.
    [0004]
    4. Method according to any one of claims 1 to 3, wherein the network is a supercomputer topology.
    [0005]
    5. A computer program product implemented on a memory medium, capable of being implemented within a computer processing unit and comprising instructions for the implementation of a method according to any one of the claims. 1 to 4.
    [0006]
    6. Switch in a network comprising a plurality of switches (af) and a plurality of nodes (0-7), this switch 15 comprising a plurality of input and output ports, able to receive and route data originating from a first node to a second node; a prerecorded list of switches directly interfaced with one or more nodes; means capable of determining the destination of said datum; means capable of selecting and then routing the data by a determined output port, when the switch is directly interfaced with the second node via said determined output port; means capable of identifying a second switch to which the second node is directly interfaced by comparison between the destination of the data and the prerecorded list of switches; means adapted to select and then route the data to the second identified switch via an output port, so that the data can be routed to said identified second switch. 35
    [0007]
    The switch according to claim 6, wherein the selection of the output port to the identified switch is performed by means capable of comparing the identified second switch with the entries of a pre-recorded routing table, each input of this table comprising - information for identifying a switch interfaced with at least one node; a set of aggregated information describing at least one route by a specific port to said switch interfaced with at least one node.
    [0008]
    8. Switch according to claim 7, comprising means for selecting for a plurality of possible routes to a switch interfaced with at least one node, an output port according to the aggregated information, this information comprising for a given port the number of jumps, congestion or weighting of links or nodes on the road to reach the switch interfaced with at least one node.
    [0009]
    9. Network of a supercomputer, comprising a plurality of switches (a-f) and a plurality of nodes (0-7), the switches being made according to any one of claims 6 to 8. B021 B005 EN Version: TQD
    类似技术:
    公开号 | 公开日 | 专利标题
    EP3155771B1|2019-11-27|Data routing method and switch for network
    US20190036810A1|2019-01-31|Method and apparatus for data network traffic control optimization
    EP3069484B1|2019-06-12|Shortening of service paths in service chains in a communications network
    US6795860B1|2004-09-21|System and method for selecting a service with dynamically changing information
    US20180219766A1|2018-08-02|Method and Apparatus for Network Traffic Control Optimization
    EP1387538B1|2007-03-21|Apparatus and method for determining routing paths in a communication network with selection attributes
    EP3189636B1|2018-11-07|Method of monitoring and of warning of routing configuration in a cluster comprising static communication links and computer program implementing this method
    FR2737624A1|1997-02-07|COMMUNICATION DEVICE AND METHOD FOR WEIGHTING AT LEAST TWO MULTI-PART COMMUNICATION LINKS BETWEEN TWO INTERLOCUTORS IN A MULTI-NODE NETWORK
    EP1432184B1|2012-04-18|Apparatus for determining communication paths in a label switching communication network having selection parameters
    FR3030168A1|2016-06-17|METHOD FOR CHOOSING AT LEAST ONE SERVICE AND ASSOCIATED DEVICE
    US10187295B2|2019-01-22|Systems and methods for routing communications within global payment networks
    US10560367B2|2020-02-11|Bidirectional constrained path search
    CN106936680B|2020-08-07|System and method for intercommunication among heterogeneous networks of cloud computing platform
    FR2903548A1|2008-01-11|METHOD FOR AUTOMATICALLY CONTROLLING A TELECOMMUNICATIONS NETWORK WITH LOCAL KNOWLEDGE MUTUALIZATION
    EP2572478B1|2017-07-19|Method of optimizing routing in a cluster comprising static communication links and computer program implementing this method
    CN109088862B|2021-01-12|Node property identification method based on distributed system
    EP3531641A1|2019-08-28|Method for establishing communication routes between nodes of a cluster of computers, corresponding computer program and cluster of computers
    EP3231137B1|2020-07-15|Overlay network for communication network connecting data centres of a cloud services provider
    EP2207313A1|2010-07-14|Method for routing a communication in a telecommunication network and associated routing management system
    Shetty et al.2016|Scalable network diversity modeling for assessing threats in cloud networks
    WO2020221779A1|2020-11-05|Methods and devices for measuring reputation in a communication network
    FR3105680A1|2021-06-25|Method and device for routing data streams
    FR3091095A1|2020-06-26|Network equipment with routing function based on an inter-network policy
    CN112055012A|2020-12-08|Distributed system
    FR3086134A1|2020-03-20|METHOD OF DETECTING ILLEGITIMATE SOURCES RESPONSIBLE FOR A DISTRIBUTED ATTACK BY DENI OF SERVICE BY LINK FLOOD AND ASSOCIATED INSTALLATION
    同族专利:
    公开号 | 公开日
    US20170126544A1|2017-05-04|
    EP3155771B1|2019-11-27|
    EP3155771A1|2017-04-19|
    US10623303B2|2020-04-14|
    WO2015193570A1|2015-12-23|
    FR3022423B1|2017-09-29|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    EP3783501A1|2019-08-23|2021-02-24|Bull SAS|Method for establishing quick communication routes between computers of a supercomputer|US7801125B2|2004-10-22|2010-09-21|Cisco Technology, Inc.|Forwarding table reduction and multipath network forwarding|
    US9054979B2|2010-04-05|2015-06-09|Maged E. Beshai|Switching system employing independent data switches connecting orthogonal sets of nodes|
    US7782882B2|2007-09-17|2010-08-24|The Boeing Company|Method and apparatus for distributing dynamic auto-summarization of internet protocol reachable addresses|
    US8611349B1|2010-06-28|2013-12-17|Amazon Technologies, Inc.|Methods and apparatus for internet-scale routing using small-scale border routers|
    US9331958B2|2012-12-31|2016-05-03|Advanced Micro Devices, Inc.|Distributed packet switching in a source routed cluster server|
    US20170012869A1|2015-07-10|2017-01-12|Microsoft Technology Licensing, Llc|Forwarding table management in computer networks|US9435658B2|2014-05-21|2016-09-06|Google Inc.|Routing with data version stitching|
    US10658847B2|2015-08-07|2020-05-19|Nucurrent, Inc.|Method of providing a single structure multi mode antenna for wireless power transmission using magnetic field coupling|
    US11205848B2|2015-08-07|2021-12-21|Nucurrent, Inc.|Method of providing a single structure multi mode antenna having a unitary body construction for wireless power transmission using magnetic field coupling|
    US10636563B2|2015-08-07|2020-04-28|Nucurrent, Inc.|Method of fabricating a single structure multi mode antenna for wireless power transmission using magnetic field coupling|
    US10985465B2|2015-08-19|2021-04-20|Nucurrent, Inc.|Multi-mode wireless antenna configurations|
    EP3504765A1|2016-08-26|2019-07-03|NuCurrent, Inc.|Wireless connector system|
    WO2018107037A1|2016-12-09|2018-06-14|Nucurrent, Inc.|A substrate configured to facilitate through-metal energy transfer via near field magnetic coupling|
    US20180233960A1|2017-02-13|2018-08-16|Nucurrent, Inc.|Method of Operating a Wireless Electrical Energy Transmission System|
    US11152151B2|2017-05-26|2021-10-19|Nucurrent, Inc.|Crossover coil structure for wireless transmission|
    US10887217B2|2018-06-29|2021-01-05|Hewlett Packard Enterprise Development Lp|Routing packets based on congestion metric thresholds and weights|
    US11271430B2|2019-07-19|2022-03-08|Nucurrent, Inc.|Wireless power transfer system with extended wireless charging range|
    US11227712B2|2019-07-19|2022-01-18|Nucurrent, Inc.|Preemptive thermal mitigation for wireless power systems|
    US11056922B1|2020-01-03|2021-07-06|Nucurrent, Inc.|Wireless power transfer system for simultaneous transfer to multiple devices|
    法律状态:
    2015-05-26| PLFP| Fee payment|Year of fee payment: 2 |
    2015-12-18| PLSC| Publication of the preliminary search report|Effective date: 20151218 |
    2016-05-26| PLFP| Fee payment|Year of fee payment: 3 |
    2017-05-23| PLFP| Fee payment|Year of fee payment: 4 |
    2018-05-25| PLFP| Fee payment|Year of fee payment: 5 |
    2020-06-26| PLFP| Fee payment|Year of fee payment: 7 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1455509A|FR3022423B1|2014-06-16|2014-06-16|METHOD OF ROUTING DATA AND SWITCH IN A NETWORK|FR1455509A| FR3022423B1|2014-06-16|2014-06-16|METHOD OF ROUTING DATA AND SWITCH IN A NETWORK|
    US15/318,885| US10623303B2|2014-06-16|2015-05-21|Method of routing data and switch in a network|
    PCT/FR2015/051344| WO2015193570A1|2014-06-16|2015-05-21|Method of routing data and switch in a network|
    EP15731640.7A| EP3155771B1|2014-06-16|2015-05-21|Data routing method and switch for network|
    [返回顶部]