Decentralized method for interconnecting N devices
A decentralized method using virtual switches in each processor to create a mesh network for satellite systems addresses the limitations of physical switches, enabling efficient communication between multiple processors with simplified configuration and robust operation.
Patent Information
- Application Number
- FR2023004248
- Authority / Receiving Office
- FR · FR
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-04-27
- Publication Date
- 2024-11-01
AI Technical Summary
Existing telecommunications solutions for interconnecting digital equipment in satellites face challenges such as limited onboard space, energy constraints, and complexity due to the need for physical switches and manual router configurations, which are not suitable for multi-processor systems.
A decentralized method using virtual switches implemented in each processor to establish a mesh network, allowing data exchange between processors via paired interfaces, eliminating the need for physical switches and simplifying configuration by defining interface pairings and implementing a neighborhood discovery protocol.
This method enables efficient communication between any two processors without physical switches, is robust to failures, and supports large numbers of interconnected processors with simplified implementation and operation, maintaining optimal routing and avoiding loops.
Smart Images

Figure 00000013_0000 
Figure 00000013_0001
Abstract
Description
Title of the invention: Decentralized method for interconnecting N devices Technical field
[0001] The invention is in the field of telecommunications between several digital devices of a telecommunications system. The invention is of particular interest in telecommunications satellites, in particular multi-processor regenerative satellites. Prior art
[0002] We consider a set of digital equipment, each comprising at least one processor, and for which it is necessary that data streams can be communicated between any two of these digital equipments, both to provide redundancy between two equipments redundant between them or more generally for the proper functioning of the applications executed on the equipments.
[0003] A classic approach to enable these data flows is to embed network switches, i.e. level 2 network hardware in the OSI model ("Open Systems Interconnection") also corresponding to the Ethernet level in the Internet nomenclature, and to connect them to the digital equipment to be interconnected. However, such a solution is difficult to reconcile with the constraints imposed on satellites in terms of limiting the number of hardware on board (to limit the number of possible breakdowns), energy power, heat dissipation and on-board space requirements.
[0004] To provide a unified telecommunications network, allowing any two digital devices in the set to communicate with each other, another solution is to assign an address per device processor telecommunications interface, to define a subnet for each pair of interfaces interconnecting two of the processors and to define routing rules for each processor. The disadvantages of this other solution are that it is complex to implement and that it requires a router (hardware or level 3 software) for each processor and therefore manual router configuration tasks to define the routes a priori. The router must be configured with all the IP addresses of the networks that it interconnects.
[0005] To avoid these constraints, point-to-point connections can be implemented. For example, with a payload of 8 processors, the point-to-point interconnection between each processor requires 7 interfaces per processor. One limitation is the number of interfaces available at the processor level, knowing that some of these interfaces may need to be redundant.
[0006] There is therefore a need for a solution for interconnecting digital equipment processors that does not have the aforementioned drawbacks. Summary of the invention
[0007] To this end, according to a first aspect, the present invention describes a method for interconnecting N digital devices, N>2, each comprising at least one processor and intended to be embedded in an electronic processing platform, each of said N processors comprising telecommunication interfaces connected to a telecommunication network, a mesh between the processors having been previously defined, by determining, relative to each of the N processors, among its interfaces, a set of interface(s) intended for the interconnection between said N processors, then pairings between interfaces of the N determined sets of interface(s), each link to a hop in the mesh corresponding to a pairing between an interface of the set of interface(s) determined for a processor and an interface of the set of interface(s) determined for another processor;
[0008] said method comprising the following steps:
[0009] iii / implementation of a virtual switch in at least some of the N processors, such a virtual switch implemented in a processor being defined to perform switching operations between the interfaces of said processor having been the subject of pairing;
[0010] iv / iv / data exchange, via said network, between two of the N processors distant from each other by two or more hops, comprising at least one switching operation, carried out by at least one of said virtual switches (20) with respect to interfaces, having been the subject of a pairing, of the processor comprising said virtual switch.
[0011] The invention thus proposes a local network solution on board a satellite, on which communications between any two processors can be implemented, not requiring a physical switch. It can be used even in digital equipment systems with a large number of processors to be interconnected, for example greater than 8.
[0012] This is a generic solution for any multi-processor edge processing.
[0013] Another advantage is its simplicity of implementation and operation: configuration only consists of listing the interfaces concerned by the virtual switch on each processor. In addition, the solution is robust to failures.
[0014] In embodiments, such a method will further comprise at least one of the following features:
[0015] - said virtual switch establishes and maintains routes between the N processors by implementation of a neighborhood discovery protocol and association of level 2 and 3 addresses relative to other processors;
[0016] - prior to steps iii and iv, the following steps are implemented:
[0017] i / relative to each of the N processors: determination, among its interfaces, of a set of interface(s) intended for the interconnection between said N processors;
[0018] ii / by pairings between interfaces of the N determined sets of interface(s), creation of a mesh between the processors, each link to a hop in the mesh corresponding to a pairing between an interface of the set of interface(s) determined for a processor and an interface of the set of interface(s) determined for another processor;
[0019] - interfaces of sets of interfaces determined for distinct processors are matched according to at least one constraint among a constraint relating to a footprint, to an availability of interfaces on each processor and in such a way that there is at least one path between all the processors taken two by two;
[0020] - the electronic processing platform is a satellite platform;
[0021] - step iii is implemented by an electronic device.
[0022] According to another aspect, the invention describes a digital processing platform, comprising N digital devices N>2, each of said N devices each comprising at least one processor comprising telecommunication interfaces connected to a telecommunication network,
[0023] a mesh between the processors having been previously defined, by determining, relative to each of the N processors, among its interfaces, a set of interface(s) intended for the interconnection between said N processors, then pairings between interfaces of the N determined sets of interface(s), each link to a hop in the mesh corresponding to a pairing between an interface of the set of interface(s) determined for a processor and an interface of the set of interface(s) determined for another processor;
[0024] at least some of the N processors each comprising a virtual switch, said switch, included in one of said processors, then being adapted to carry out switching operations between the interfaces of said processor which have been the subject of pairing;
[0025] wherein two of the N processors two or more hops apart from each other are adapted to exchange data between them, via said network, and at least one of said N virtual switches is adapted to contribute to implementing said exchange by performing at least one switching operation, with respect to interfaces having been the subject of a pairing, of the processor comprising said virtual switch.
[0026] In embodiments, such a platform will further comprise at least one of the following features:
[0027] - a virtual switch is adapted to establish and maintain routes between the N processors by implementing a neighborhood discovery protocol and associating level 2 and 3 addresses relative to other processors;
[0028] - interfaces of sets of interfaces determined for distinct processors have been matched according to at least one constraint among a constraint relating to a footprint, to an availability of interfaces on each processor and in such a way that there is at least one path between all the processors taken two by two;
[0029] - the electronic processing platform is a satellite platform. Brief description of the drawings
[0030] The invention will be better understood and other characteristics, details and advantages will appear more clearly on reading the following description, given without limitation, and thanks to the appended figures, given by way of example.
[0031] [Fig-1] [Fig.l] is a representation of a set of processors of interconnected digital equipment in one embodiment of the invention;
[0032] [Fig.2] [Fig.2] represents the steps of a method of interconnecting a set of digital equipment processors in one embodiment of the invention.
[0033] Identical references may be used in different figures when they designate identical or comparable elements. Description of the embodiments
[0034] With reference to [Fig.l], we consider a set 1 of N separate digital devices on board or intended to be on board a telecommunications satellite (not shown), for example a regenerative satellite. N is greater than or equal to 2. These digital devices are for example signal processing devices, in particular digital formation of radiocommunication beams, packet routing, monitoring / control of software executed on a processor or specific processing on data packets (eg concatenation, re-sequencing).
[0035] Here, as an example, N has been chosen equal to 8. Each of these N devices comprises at least one processor 10: the ith device, i = 1 to N, thus comprises the processor 10 named Proc i.
[0036] Each device comprises a memory in which software applications are stored and is adapted to execute each of these software applications on its processor. At least some of the applications generate data exchanges between processors 10. In one embodiment, some of the applications are distributed between several devices.
[0037] Each of the processors 10 is connected to a local telecommunications network, for example wired or wireless, for example radio.
[0038] Each of the processors 10 comprises one or more telecommunications interfaces connected to this local network. These interfaces are Ethernet interfaces and correspond to separate physical connectors.
[0039] [Fig.2] represents the steps of a method of interconnecting the processors 10 of the digital equipment of the set 1 in one embodiment of the invention.
[0040] In a prior step 101, relative to each processor, the identification of one or more of its available interfaces to be dedicated to the implementation of the interconnection between the processors is carried out: the choice of the interfaces concerned is defined by the designer when defining the local network.
[0041] A mesh of the processors 10 is then produced using the interfaces thus identified, in which interfaces 11 of distinct processors 10 are paired with each other, knowing that an interface of a processor can only be paired with a single other interface and that this single other interface must be that of a single other processor 10. Two interfaces 11 thus paired with each other thus create a connection (to a hop) of the mesh between the processors 10 to which these interfaces belong. The mesh produced via these interface pairings must allow any processor 10 to be connected to any other processor 10 by at least one connection comprising one or more successive hops of the mesh.
[0042] Thus below, when it is mentioned “interfaces of the same processor which were paired in step 101”, this refers to interfaces of the same processor each paired with interfaces of another processor (and not to interfaces of the same processor which would be paired with each other).
[0043] Depending on the embodiments, additional constraints are taken into account for the definition of the mesh: a maximum number of hops to be respected for a connection between two processors or a maximum number of interfaces used for the pairing in total and / or per processor (or for certain processors at least).
[0044] In embodiments, in order to limit the number of interfaces used per processor, there are one or more processors not directly connected (via one of its interfaces) to each other processor, i.e. the data transmitted between these two processors will have to pass through at least one other processor. In one embodiment, only a strict subset of the interfaces of a processor are subject to pairing.
[0045] The mesh of the processors 10 obtained in the embodiment considered is represented in [Fig.l]:
[0046] - thus, in the processor Proc 1: the interface 11 named Int 1.3 is paired with interface 11 named Int 4.2 of processor Proc 4; interface 11 named Int 1.1 is paired with interface 11 named Int 5.1 of processor Proc 5; interface 11 named Int 1.2 is paired with interface 11 named Int 2.1 of processor Proc 2;
[0047] - in processor Proc 2: interface Int 2.1 is paired with interface Int 1.2 of the processor Proc 1; interface Int 2.2. is paired with interface Int 3.2 of processor Proc 3;
[0048] - etc.
[0049] The connections to a hop (here 9 in number) corresponding to the paired interfaces are referenced 12 in [Fig.l].
[0050] Once the interconnection graph of the processors 10 has been thus defined, in a step 102, a virtual switch 20 is implemented, in each processor 10: it is instantiated, by exploiting (only) those of the processor interfaces which have been paired with interfaces of other processors 10 in step 101. These interfaces are thus specifically associated with the virtual switch during the instantiation, to the exclusion of those which have not been associated with it: for example the interfaces Int 5.1, Int 5.2 for the vCom switch 5 (and not Int 5.3 or any other available interfaces of the Proc 5 but not paired); during its operational operation, the vCom virtual switch 5 will not monitor the Int 5.3 interface and will not route data received on this interface or to be transmitted on this interface.
[0051] Virtual switch 20 is a program running on edge processor 10 intercepting traffic to and from interfaces of the processor that were paired with interfaces of other processors 10 in step 101.
[0052] In a Linux-based example, the instantiation of the Proc 5 processor includes the following commands: - to create and mount the virtual instance of the vCom 5 switch: ip link add name vCom 5 type bridge stp_state 1
[0053] ip link set vCom 5 up - to define and associate with vCom 5 the only processor interfaces concerned by the operations of the vCom 5 virtual switch, i.e. here the int51 and int52 interfaces: ip link set dev int51 master vCom 5
[0054] ip link set dev int52 master vCom 5
[0055] These elements can also be entered in a simple configuration file, for example using a network configuration tool such as netplan® for deployment.
[0056] The Spanning Tree® Protocol (STP) is for example used by virtual switches for the identification and implementation of the routing strategy automatically; thus virtual switches, for example by implementing STP work between them: 1. identifies the available routes in the interconnection network corresponding to the mesh; 2. defines the nominal and redundant routes from any processor 10 to another processor 10 and 3. establishes the nominal route.
[0057] Using a Layer 2 approach can lead to unnecessary loops and data streaming: STP prevents these side effects from occurring.
[0058] In one embodiment, a virtual switch implements the ARP protocol. In particular, when it has to send data to a level 3 address, hereinafter called "address B", to which it has not yet associated a level 2 address, it sends ARP requests to this address in broadcast, i.e. to all the virtual switches using a predefined generic Ethernet destination address. Such an ARP ("Address Resolution Protocol") request asks who has address B. The one who has address B responds by transmitting its MAC address (level 2).
[0059] The invention can be illustrated by the exploitation of these commands, here Linux, without a centralized switch.
[0060] Of course, other types of instantiation can be used, for example by using other operating systems, for example network software from network equipment suppliers such as CISCO ® or JUNIPER ®.
[0061] The processor Proc i thus comprises a virtual switch vCom i, i = 1 to N, operating at level 2, which behaves like a physical switch.
[0062] The virtual switch within a processor therefore performs the routing / switching tasks on data passing through the interfaces of the processor that have been paired (data entering through an interface of the processor that has been paired may be destined for - or originating from - an application of the processor or may simply pass through the processor acting as a relay via the virtual switch). The network of virtual switches thus created at level 2 provides the same service as that which would be provided by a physical switch directly connected to each of the processors 10.
[0063] The virtual switch within a processor does not perform routing / switching on data passing through any interface of the processor that was not paired in step 101.
[0064] The invention does not use on-board and centralized router or switch equipment, interconnecting the network elements. Through this solution, each processor hosts part of the elements necessary for establishing communication between each of the processors even if they are not all directly connected to each other. Each processor discovers the topology from near to far and contributes to defining optimal routing between processors, for example avoiding data duplication and loops.
[0065] The switch vCom i is virtual, that is to say software: its functions are implemented by execution on the processor Proc i of software instructions stored in a memory of the processor Proc i.
[0066] Then, to exploit the proposed solution, the following approaches are for example implemented: - a level 3 address (typically an IP address), static or dynamic, is assigned respectively to each virtual switch 20; - a unique name is associated with each processor Proc i, i = 1 to N; - higher level services, using level 3 addresses or the name single processor, can thus access applications running on any of the 10 processors without a priori knowledge of the network topology.
[0067] Steps 101 and 102 are carried out according to the embodiments automatically, for example by an electronic interconnection implementation device or manually.
[0068] In a step 103, during the operational functioning of the processors 10, data streams are exchanged between the processors 10. The data is routed from a source processor 10 to a destination processor 10, exclusively on the connections of the mesh and in a manner controlled by the virtual switches. The processors 10 are directly connected to each other, via their paired interfaces.
[0069] A virtual switch 20 in a processor 10 and having to route, to a defined recipient, a data packet received on one of the interfaces 11 (let us call it the 'interface 11R') associated with the virtual switch thus performs a routing operation with respect to the interfaces of said processor which have been the subject of a pairing: it determines which of the associated interfaces the packet must be delivered to (let us call it the 'interface 11D') as well as the route to be followed by this packet.
[0070] This virtual switch thus contributes to implementing the exchange between the destination processor and the source processor of the packet, for example two hops (or more) apart, by carrying out at least this routing operation between the two interfaces 11R and 11D of the processor comprising this virtual switch.
[0071] Thus a software application (level 3, i.e. IP layer) running on the processor Proc 1 will be able to communicate with an application of the same level running on the processor Proc 8 without needing to know / manage that the data must pass through intermediate processors. The invention therefore makes it possible to obtain a unified network, allowing each processor to communicate with any other processor in the set, without a network switch and while there is no necessarily a direct connection between each of the processors. The solution allows direct communication at the application level without a direct connection between the processors.
[0072] The invention corresponds to a decentralization of the routing / switching functions (performed in the prior art via physical routers or switches) allowing interconnection between the different processors: the virtual switch within a processor performs the routing tasks on data transiting via any interface of the processor that has been paired.
[0073] The invention is simple to apply, both for implementation and for subsequent exploitation, even for high values of N.
Claims
Claims
1. Method for interconnecting N digital devices, N>2, each comprising at least one processor (10) and intended to be embedded in an electronic processing platform (1), each of said N processors (10) comprising telecommunication interfaces (11) connected to a telecommunication network, a mesh between the processors (10) having been previously defined, by determining, relative to each of the N processors, among its interfaces (11), a set of interface(s) intended for the interconnection between said N processors, then pairings between interfaces of the N determined sets of interface(s), each link to a hop (12) in the mesh corresponding to a pairing between an interface (11) of the set of interface(s) determined for a processor and an interface (11) of the set of interface(s) determined for another processor;said method comprising the following steps: iii / implementing a virtual switch (20) in at least some of the N processors (10), such a virtual switch (20) implemented in a processor (10) being defined to perform switching operations between the interfaces of said processor having been the subject of a pairing; iv / exchanging data, via said network, between two of the N processors distant from each other by two or more hops, comprising at least one switching operation, performed by at least one of said virtual switches (20) with respect to interfaces, having been the subject of a pairing, of the processor comprising said virtual switch.;
2. A method of interconnecting N digital devices according to claim 1, wherein said virtual switch (20) establishes and maintains routes between the N processors (10) by implementing a neighborhood discovery protocol and associating level 2 and 3 addresses relating to the other processors (10).
3. Method for interconnecting N digital devices according to claim 1, comprising, prior to steps iii and iv, the following steps: - i / relative to each of the N processors (10): determination, among its interfaces (11), of a set of interface(s) intended for the interconnection between said N processors; - ii / by pairings between interfaces of the N determined sets of interface(s), creation of a mesh between the processors, each link to a hop in the mesh corresponding to a pairing between an interface (11) of the set of interface(s) determined for a processor and an interface of the set of interface(s) determined for another processor.
4. Method for interconnecting N digital devices according to claim 3, according to which interfaces (11) of sets of interfaces determined for distinct processors (10) are paired according to at least one constraint among a constraint relating to a size, to an availability of interfaces on each processor and in such a way that there is at least one path between all the processors taken two by two.
5. Method for interconnecting N digital devices according to any one of the preceding claims, the electronic processing platform (1) being a satellite platform.
6. Method for interconnecting N digital devices according to any one of the preceding claims, according to which step iii is implemented by an electronic device.
7. Digital processing platform (1), comprising N digital devices N>2, each of said N devices (10) each comprising at least one processor (10) comprising telecommunication interfaces (11) connected to a telecommunication network, a mesh between the processors (10) having been previously defined, by determining, relative to each of the N processors, among its interfaces (11), a set of interface(s) intended for the interconnection between said N processors, then pairings between interfaces of the N determined sets of interface(s), each link to a hop (12) in the mesh corresponding to a pairing between an interface (11) of the set of interface(s) determined for a processor and an interface (11) of the set of interface(s) determined for another processor; at least some of the N processors (10) each comprising a virtual switch (20), said switch, included in one of said pro- processors, then being adapted to perform switching operations between the interfaces of said paired processor; wherein two of the N processors two or more hops apart from each other are adapted to exchange data between them, via said network, and at least one of said N virtual switches (20) is adapted to contribute to implementing said exchange by performing at least one switching operation with respect to interfaces (11), paired, of the processor comprising said virtual switch.
8. A digital processing platform according to claim 7, wherein a virtual switch (20) is adapted to establish and maintain routes between the N processors (10) by implementing a neighborhood discovery protocol and associating level 2 and 3 addresses relating to the other processors (10).
9. Digital processing platform according to any one of claims 7 or 8, in which interfaces (11) of sets of interfaces determined for distinct processors (10) have been paired according to at least one constraint among a constraint relating to a size, to an availability of interfaces on each processor and in such a way that there is at least one path between all the processors taken two by two.
10. A digital processing platform according to any one of claims 7 to 9, the electronic processing platform (1) being a satellite platform.
Citation Information
Patent Citations
Address resolution protocol system and method in a virtual network
US20020156612A1
Infrastructure processing unit
US20210117242A1
Virtual networking system and method in a processing system
WO2002086712A1
