Virtualization architecture construction method for large-scale dynamic routing simulation system

By starting multiple virtual machines on a single physical server and cutting network topology, building a virtualized architecture, the problem of difficulty in simulating large-scale dynamic routing systems in the existing technology is solved, and the simulation effect of efficient parallel processing and resource conservation is achieved.

CN119030924AInactive Publication Date: 2024-11-26COMP NETWORK INFORMATION CENT CHINESE ACADEMY OF SCI

Patent Information

Application Number
CN202411175144.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-26
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing routing simulation technologies are difficult to meet the simulation needs of large-scale dynamic routing systems. Hardware-based methods are costly and inflexible, and software-based methods are difficult to handle a large number of links and routing changes in parallel.

Method used

Using a virtualization architecture method for large-scale dynamic routing simulation systems, by starting multiple virtual machines on a physical server, cutting the network topology into multiple subgraphs, each virtual organization builds a virtual network topology, and using containers or namespaces Encapsulate the routing protocol module to implement virtual links and VPN tunnels across sub-graph edges to ensure efficient resource utilization and parallel processing capabilities.

Benefits of technology

It realizes efficient simulation of large-scale routing systems on a single physical server, reduces resource consumption and simulation time, improves the accuracy and efficiency of simulation results, and fills the gaps in the existing technology.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119030924A_ABST
    Figure CN119030924A_ABST
Patent Text Reader

Abstract

The invention discloses a virtualization architecture construction method for a large-scale dynamic routing simulation system, which comprises the following steps of: starting a plurality of virtual machines which are connected by using an under lay network in any form; cutting the network topology to be simulated into a plurality of sub-graphs of which the number is the same as that of the virtual machines, and constructing a virtual network topology corresponding to one sub-graph on each virtual machine; in each virtual machine, virtualizing each node in the subgraph into a routing protocol module packaged by using a container or a namespace; a virtual link is achieved through the inner edge of each sub-graph in the sub-graphs; a VPN tunnel is formed by each cross-sub-graph edge, connected with other sub-graphs, in the sub-graphs, and a globally unique tunnel ID number is distributed to the tunnel. The routing simulation platform based on the network virtualization technology overcomes the defects that an existing routing simulation platform based on the network virtualization technology is too large in resource overhead or difficult to support a large number of link changes and routing changes, and the routing simulation platform can meet the simulation requirement of a large-scale routing system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to routing simulation and network virtualization technology, and in particular to a virtualization architecture method for a large-scale dynamic routing simulation system. Background Art

[0002] Routing simulation is an important technology to promote the development of network technology. By building a virtual routing system, operators or researchers can deepen their understanding of current network routing technology, identify problems with current network routing technology, and verify new network routing technology. At present, routing simulation has been widely used in teaching, attack and defense drills, new network technology experiments and other scenarios, and has actually promoted the development of network technology.

[0003] The research on hot network issues and the development of new network technologies have given rise to the demand for large-scale routing simulation. For example, the study of global BGP routing security requires an experimental platform that can simulate tens of thousands of AS nodes; the development of 5G network technology requires an experimental platform that can simulate a large number of routing updates caused by users going online and offline; the study of new low-orbit satellite network technology requires an experimental platform that can simulate thousands to tens of thousands of satellite routing nodes, as well as a large number of routing updates caused by link changes. The characteristics of the network simulated by the above new requirements are large-scale and dynamic, that is, the node scale is large, and there are a large number of frequent link changes and routing changes. However, the existing routing simulation technology routes are difficult to meet the above large-scale routing simulation needs.

[0004] The existing routing simulation technology routes can be roughly divided into two types: hardware-based and software-based. The hardware-based routing simulation platform connects multiple real routing devices through links to build a hardware test bed for operators to use. There are two implementation schemes for software-based routing simulation: 1) Use simulation software based on discrete event models, such as NS3, OMNet++, etc.; 2) Build a virtual network based on network virtualization technology, that is, use isolation mechanisms such as virtual machines, containers or namespaces to build virtual routing nodes, and connect them through virtual links. For example, "A Space-Earth Fusion Network Simulation Test System and Its Test Method" (CN114785401A) uses the docker isolation mechanism to build virtual satellite network nodes, uses the veth interface and the Linux bridge to connect each virtual node, and uses the Linux tc tool to dynamically control the link, and builds a routing simulation platform for satellite networks. "A Mobile Wireless Network Simulation and Experimental Platform Based on Virtualization" (CN109039709A) uses virtual machines to build virtual routing nodes, and connects each virtual routing node through virtual links based on Linux kernel forwarding, building a simulation platform for wireless mobile network teaching. In addition, the above-mentioned software-based simulation technology route can be connected with real hardware to form a semi-physical routing simulation system combining software and hardware, so that the simulation results are more in line with the actual situation. For example, "Method and System for Constructing a Virtualized Network Security Range in a Power Plant" (CN116668295A) builds a virtual network based on Openstack virtualization technology and connects it to a physical router to build a power plant network security test range.

[0005] The above-mentioned existing technologies are difficult to simulate large-scale routing systems. Building a large-scale routing system test bed based on hardware requires a large number of hardware devices, which leads to high costs, difficulty in operation and poor flexibility. The simulation platform based on the discrete event model requires the scheduler to reschedule the event queue when processing events, so it is difficult to quickly and in parallel process a large number of link changes / route changes / data packet sending and receiving events generated by large-scale routing systems during operation; the routing simulation platform based on network virtualization technology is difficult to simulate large-scale routing systems for two reasons: the virtual single routing node consumes too many resources and it is difficult to support the operation of large-scale virtual routing nodes; multiple link changes, route changes and other events are difficult to process in parallel, and it is difficult to support the rapid processing of a large number of link changes and route changes. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of a routing simulation platform based on network virtualization technology and enable the routing simulation platform to meet the simulation requirements of a large-scale routing system.

[0007] To achieve the above object, the present invention provides a method for constructing a virtualized architecture for a large-scale dynamic routing simulation system, the method comprising:

[0008] Start multiple virtual machines, each of which is connected using any form of underlay network;

[0009] The network topology of the routing system to be simulated is divided into multiple subgraphs equal to the number of virtual machines, and a virtual network topology corresponding to a subgraph is constructed on each virtual machine;

[0010] In each virtual machine, each node in the above subgraph is virtualized into a routing protocol module encapsulated by a container or namespace; each internal edge in the above subgraph is implemented as a virtual link; each cross-subgraph edge in the above subgraph connected to other subgraphs is implemented as a VPN tunnel, and a globally unique tunnel ID number is assigned to the tunnel.

[0011] The present invention overcomes the shortcomings of existing routing simulation platforms based on network virtualization technology, such as either excessive resource overhead or difficulty in supporting a large number of link changes and routing changes, so that the routing simulation platform can meet the simulation requirements of large-scale routing systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic diagram of a routing simulation architecture using a virtual machine for the prior art;

[0013] Figure 2 Schematic diagram of routing simulation architecture using lightweight isolation mechanism for existing technologies;

[0014] Figure 3 A schematic diagram of a virtualization architecture for a large-scale dynamic routing simulation system provided by an embodiment of the present invention;

[0015] Figure 4 It is a schematic diagram of the topological segmentation effect of an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] To facilitate understanding of the embodiments of the present invention, specific embodiments will be further explained below in conjunction with the accompanying drawings. The embodiments do not constitute a limitation on the embodiments of the present invention.

[0018] The existing routing simulation platform based on network virtualization technology needs to complete two aspects of link virtualization and routing node virtualization. The technologies for implementing virtual links are basically the same. They often use virtual links based on the Linux kernel, such as Linux bridges, OVS switches, etc., to connect various virtual routing nodes. The overall idea of ​​implementing virtual routing nodes is to encapsulate routing protocol modules through virtualization technology to achieve isolation between multiple routing protocol modules. The routing protocol module performs routing calculation, routing message sending and receiving, and routing delivery. Both routing message sending and receiving and routing delivery need to interact with the network module of the Linux kernel. There are two solutions for isolating virtual routing nodes, each with its own defects.

[0019] One is to use virtual machines for isolation. Figure 1 As shown in the figure, on a physical server, multiple virtual machines are started, and the above-mentioned routing protocol module is run inside each virtual machine. The main advantage of this technical route is that different routing protocol modules and virtual links do not share the kernel, and the interactive operation with the Linux network module is not subject to the kernel global lock, so it can be parallel, and a large number of link changes and routing delivery operations can be completed in a short time. The main defect of this technical route is that a single virtual machine contains a system kernel, which causes it to occupy too much memory resources. Therefore, a large number of virtual machines cannot be started on a physical server, and large-scale virtual routing nodes cannot be simulated. Simulation can be performed by using a cluster containing multiple physical servers, but the cost is too high.

[0020] The second is to use lightweight isolation mechanisms such as containers or namespaces for isolation. Figure 2As shown, on a physical server, multiple containers or network namespaces are opened, and the above-mentioned routing protocol module is run inside each container or network namespace. The main advantage of this technical route is that the memory overhead brought by each container or network namespace is small, a large number of containers or network namespaces can be opened on a physical server, and large-scale virtual routing nodes can be simulated. The main defect of this technical route is that all routing protocol modules and virtual links share the kernel, and the interactive operations with the Linux network module are subject to the same kernel global lock and cannot be parallel. The specific manifestation of the above defect is: when multiple link change events occur, it is necessary to initiate RTM_GETLINK and RTM_NEWTFILTER and other operation requests to the Linux kernel through the net link message to change the link state, but the above RTM_GETLINK operation is in the critical section of the kernel global lock rtnl_mutex and cannot be parallel; when multiple routing change events occur, it is necessary to initiate RTM_ADDROUTE and RTM_DELROUTE and other operation requests to the Linux kernel through the net link message, but the above operations are also in the critical section of the kernel global lock rtnl_mutex and cannot be parallel. The two major consequences of the above-mentioned link changes and routing changes being unable to be processed in parallel are: the simulation progresses slowly; the routing changes cannot be processed in parallel, resulting in the routing convergence logic in the simulation being different from that in the real routing system, and the simulated routing convergence time being inaccurate. Therefore, the above-mentioned routing simulation architecture using a lightweight isolation mechanism is difficult to meet the needs of large-scale routing system simulation.

[0021] The embodiment of the present invention provides a method for constructing a virtualized architecture for a large-scale dynamic routing simulation system. Figure 1 As shown, on a physical server, first start multiple virtual machines, each virtual machine is connected using any form of underlay network (in the figure, they are connected using virtual switches); then the network topology to be simulated is cut into multiple subgraphs equal to the number of virtual machines, and a virtual network topology corresponding to a subgraph is constructed on each virtual machine. In each virtual machine, each node in the above subgraph is virtualized into a routing protocol module encapsulated using a container or namespace, that is, a virtual routing node; each subgraph internal edge in the above subgraph is implemented as a virtual link built based on the Linux kernel network module; each cross-subgraph edge in the above subgraph connected to other subgraphs is implemented as a VPN tunnel, and a globally unique tunnel ID number is assigned to the tunnel. The above cross-subgraph virtual link establishment method unifies multiple cross-subgraph virtual links to the same VPN tunnel endpoint, which is easy to manage and ensures the isolation of each virtual link.

[0022] The embodiment of the present invention combines Figure 1 and Figure 2The advantages of the two technical architectures complement each other. On a physical server, it is only necessary to start a small number of virtual machines to appropriately increase the parallelism of link change and route change processing, and ensure that the above-mentioned change events within the time slice can be processed within a simulation step time slice. In other words, the embodiment of the present invention requires that the number of virtual machines started is much less than the number of routing nodes of the simulated large-scale routing system. Therefore, each change event can be quickly processed without occupying too many memory resources, ensuring the advancement speed of the simulation and making the simulated routing convergence as close to the actual situation as possible. Therefore, the embodiment of the present invention can meet the simulation needs of large-scale routing systems and fill the gap in existing routing simulation work.

[0023] The embodiment of the present invention optimizes the virtualization architecture of only a single physical server. When multiple servers are used to expand the simulation platform, the virtualization architecture proposed in this solution can be run on each physical server.

[0024] In a specific embodiment, Figure 3 As shown in the figure, the construction process on a single physical server is as follows:

[0025] Step 1: Receive user input.

[0026] The content that the user needs to enter includes: the number of virtual machines N_vm started on a single physical server.

[0027] The initial network topology of the simulated routing system must include: routing node information: a list containing all m nodes in the topology: [node_1, node_2, ... node_m], each node is a network node with routing function.

[0028] Initial link information: a set of all n edges in the topology graph: {edge_1, edge_2, ..., edge_n}, where each edge edge_k is a tuple consisting of the two vertices of the edge<node_p,node_q> .

[0029] The link changes and route change information of the simulated routing system include: Link change information: a list of link change events, such as link interruption and recovery caused by laser vibration and solar eclipse when simulating highly dynamic satellite Internet; Route change information: a list of route change events specified by the user; It should be noted that during the simulation process, link changes will be sensed by the upper-layer routing protocol module, triggering a certain degree of route changes. During the simulation process, the route changes of each virtual routing node are jointly determined by the route changes triggered by the above link changes and the above user-specified route changes.

[0030] Step 2: Segment the network topology.

[0031] Divide the initial network topology graph in step 0 into several subgraphs. The specific steps are as follows:

[0032] Split the node list. Split the node list corresponding to the routing node information in step 0 into N_vm segments in order, forming N_vm child node lists (when m cannot divide N_vm, let the length of the 1st to (N_vm-1)th child node lists be int(m / N_vm), and let the length of the N_vmth child node list be [m-int(m / N_vm)*(N_vm-1)]).

[0033] Split the edge list. Split the edge list corresponding to the initial link information in step 0 into 2*N_vm sets, where the first N_vm sets are subgraph internal edge sets and the last N_vm sets are cross-subgraph edge sets.

[0034] The specific steps to construct the above set are:

[0035] (1) Initialize each of the above sets to an empty set;

[0036] For each edge edge_k in the edge set corresponding to the initial link information in step 0:<node_p,node_q> , let node_p be included in the list of child nodes in step 0 of the i-th step, and node_q be included in the list of child nodes of the j-th step. If i == j, add edge_k to the internal edge set of the i-th subgraph; if i ! = j, add<node_p,node_q> Add the i-th cross-subgraph edge set,<node_q,node_p> Add the jth cross-subgraph edge set.

[0037] After the above steps are completed, the topology corresponding to the t-th subgraph can be determined by the t-th child node list, the t-th subgraph internal edge set, and the t-th cross-subgraph edge set. Figure 4 shown.

[0038] In addition to the above methods, users can customize other topological segmentation methods.

[0039] Step 3: Based on virtual machine technology, build the underlying virtualization environment of multiple virtual machines. The specific steps are as follows:

[0040] On a physical server, start multiple virtual machines;

[0041] Connect the above multiple virtual machines using any form of underlay network. Underlay networks include but are not limited to the following solutions: virtual network ports with virtual switches; use SR-IOV network card VF to connect each virtual machine;

[0042] The purpose of step 0 is to create a simulation platform environment containing multiple core global locks so that link change and route change events during the simulation process can be processed quickly and in parallel.

[0043] Based on the lightweight isolation mechanism, an upper-layer virtualized network environment is established. In each virtual machine, virtual routing nodes and virtual links are run based on the network topology corresponding to the subgraph described in step 0. Assuming that the subgraph is determined by the subnode list NodeList, the subgraph internal edge set IntraEdgeList, and the cross-subgraph edge set InterEdgeList, the specific steps of running the virtual routing nodes and virtual links are as follows:

[0044] For each node_j in NodeList, create a lightweight isolation unit, which can be implemented as a container or a network namespace;

[0045] For each link in IntraEdgeList<node_p,node_q> , create a virtual link to connect the lightweight isolation units corresponding to node_p and node_q. ​​The virtual link can be implemented based on the Linux kernel network module;

[0046] For each link in InterEdgeList<node_p,node_q> , create a VPN tunnel endpoint, bind the VPN tunnel to the port leading to the underlay network between virtual machines described in step 0, and assign a globally unique tunnel ID number to the VPN tunnel. Since node_q runs on another virtual machine, a VPN tunnel peer endpoint will be created on the other virtual machine. After both endpoints are created, the VPN tunnel is connected, and a cross-subgraph virtual link is established.

[0047] The purpose of step 0 is to use lightweight virtualization technology to build a virtual network environment based on the network topology of the routing system to be simulated, thereby saving resources consumed in building the virtual network topology.

[0048] In each lightweight isolation unit of each of the above virtual machines, a routing protocol module is run to start simulating routing functions such as routing calculation, routing message sending and receiving, and routing delivery.

[0049] After executing the above process and building a virtual network based on a multi-level hybrid virtualization architecture, you can start the routing simulation promotion process to dynamically adjust the link and routing status within the virtual network based on the input link change and routing change information.

[0050] The above process is only built for the virtualization architecture of a single physical server. When using multiple servers to expand the simulation platform, the above process can be executed on each physical server.

[0051] The embodiments of the present invention overcome the shortcomings of existing routing simulation platforms based on network virtualization technology, such as either excessive resource overhead or difficulty in supporting a large number of link changes and routing changes, so that the routing simulation platform can meet the simulation requirements of large-scale routing systems.

[0052] It is obvious that many variations of the invention described herein are possible without departing from the true spirit and scope of the invention. Therefore, all modifications obvious to those skilled in the art should be included within the scope of the claims. The scope of the invention is limited only by the claims.

Claims

1. A method for constructing a virtualized architecture for a large-scale dynamic routing simulation system. It is characterized in that include: Start multiple virtual machines, each of which is connected using any form of underlay network; The network topology of the routing system to be simulated is divided into multiple subgraphs equal to the number of virtual machines, and a virtual network topology corresponding to a subgraph is constructed on each virtual machine; In each virtual machine, each node in the above subgraph is virtualized into a routing protocol module encapsulated by a container or namespace; each internal edge in the above subgraph is implemented as a virtual link; each cross-subgraph edge in the above subgraph connected to other subgraphs is implemented as a VPN tunnel, and a globally unique tunnel ID number is assigned to the tunnel.

2. The method according to claim 1, It is characterized in that The network topology of the routing system to be simulated, including: Routing node information: contains a list of all m nodes in the topology graph, each of which is a network node with routing function; Initial link information: contains the set of all n edges in the topology graph, where each edge is a tuple consisting of its two vertices.

3. The method according to claim 1, It is characterized in that Link change and route change information of the routing system to be simulated, including: Link change information: contains a list of link change events, including link interruption and recovery caused by laser vibration and solar eclipse when simulating highly dynamic satellite Internet; Route change information: contains a list of route change events specified by the user; During the simulation process, link changes will be sensed by the upper-layer routing protocol module, triggering routing changes; during the simulation process, routing changes of each virtual routing node are jointly determined by routing changes triggered by the above link changes and routing changes specified by the above users.

4. The method according to claim 2, It is characterized in that The step of cutting the network topology of the routing system to be simulated includes: Split the node list, and evenly split the node list corresponding to the routing node information into N_vm segments in order to form N_vm child node lists; Split the edge list, and split the edge list corresponding to the initial link information into 2*N_vm sets, where the first N_vm sets are subgraph internal edge sets, and the last N_vm sets are cross-subgraph edge sets.

5. The method according to claim 4, It is characterized in that The steps to build a collection include: Initialize each set to an empty set; For each edge in the edge set corresponding to the initial link information, let node_p be included in the i-th child node list and node_q be included in the j-th child node list; if i == j, add each edge to the i-th subgraph internal edge set; if i != j, add<node_p,node_q> Add the i-th cross-subgraph edge set,<node_q,node_p> Add the jth cross-subgraph edge set; After the above steps are executed, the topology corresponding to the t-th subgraph can be determined by the t-th child node list, the t-th subgraph internal edge set, and the t-th cross-subgraph edge set.

6. The method according to claim 1, It is characterized in that In each virtual machine, virtual routing nodes and virtual links are run based on the network topology corresponding to the subgraph; the subgraph is assumed to be determined by a subnode list, a subgraph internal edge set, and a cross-subgraph edge set.

7. The method according to claim 6, It is characterized in that The steps to run a virtual routing node and a virtual link include: For each node_j in the child node list, create a lightweight isolation unit, which is implemented as a container or a network namespace; For each link in the subgraph's internal edge set<node_p,node_q> , create a virtual link to connect the lightweight isolation units corresponding to node_p and node_q; For each link in the set of cross-subgraph edges<node_p,node_q> , create a VPN tunnel endpoint, bind the VPN tunnel to the port leading to the under lay network between virtual machines, and assign a globally unique tunnel ID number to the VPN tunnel; because node_q runs on another virtual machine, a VPN tunnel peer endpoint will be created on the other virtual machine; after both endpoints are created, the VPN tunnel is connected, and a cross-subgraph virtual link is established.

8. The method according to claim 1, It is characterized in that In the user input step, the content that the user needs to input includes: the number of virtual machines N_vm started on a single physical server, and link change and route change information.

9. The method according to claim 1, It is characterized in that The under lay network includes: virtual network ports with virtual switches, and VFs using SR-IOV network cards to connect each virtual machine.

10. The method according to claim 1, It is characterized in that Virtual Link is built on top of the Linux kernel network module.

Citation Information

Patent Citations

  • SDN-based virtual network topology automatic construction method

    CN115118607A

  • Container-based routing simulation system and method

    CN118250180A

  • Method and device for realizing network slicing, and controller

    WO2020103902A1

Cited By

  • Method and system for simulating VPN (Virtual Private Network) access in network target range

    CN119583366A

  • Routing network-oriented virtual machine configuration and node allocation method and device

    CN121814681A