A method for partial deployment of SRv6 node selection and traffic planning based on celestial balance

Abstract

This invention relates to a method for selecting and planning traffic for partially deployed SRv6 nodes based on celestial equilibrium, belonging to the field of computer network technology. The method includes: analogizing the influence of SRv6 nodes to the gravitational pull of a star on surrounding planets in celestial equilibrium theory; calculating network load based on the influence area of ​​each SRv6 node; using an iterative algorithm to select the SRv6 node and its influence range that currently provides the best network load balancing effect; repeating the iteration, analogous to the celestial equilibrium process, until the distribution of SRv6 nodes and their influence areas converges; and performing traffic planning based on celestial equilibrium theory, constructing the optimal routing path for each traffic in the current traffic set. This invention optimizes the distribution of SRv6 nodes based on celestial equilibrium theory, comprehensively considering the global network topology and traffic distribution, and does not limit each traffic to only two routing segments. Using this method, a better network load balancing effect can be achieved.

Description

Technical Field

[0001] This invention belongs to the field of computer network technology, specifically relating to a method for partial deployment of SRv6 nodes and traffic planning based on celestial balance. Background Technology

[0002] SRv6, or Segment Routing over IPv6, is an emerging network technology. SRv6 is an extension of Segment Routing (SR) technology to IPv6. By adding SR tags to IPv6 packets, traffic can be routed to different SR segments, allowing for dynamic adjustment of traffic paths and effectively achieving traffic distribution, ensuring a balanced load across all links. This feature makes SRv6 a promising candidate for load balancing. Network devices implementing SR technology are called SR nodes. However, upgrading all nodes in a network to SR nodes at once is impractical due to prohibitive costs. Therefore, selecting the optimal nodes to upgrade to SR nodes within a limited number of upgradable nodes to optimize load balancing is a critical and complex issue; this approach is known as partially deployed SR-TE (Segment Routing Traffic Engineering).

[0003] Partial deployment of SR-TE involves selecting only a subset of nodes in the network as SR nodes. The work required can be divided into two key aspects: selecting appropriate SR nodes and planning traffic paths. After selecting suitable SR nodes, it is also necessary to decide which intermediate SR nodes each flow in the incoming traffic set should pass through in order to distribute traffic and achieve the goal of load balancing.

[0004] Current research methods for deploying SR-TE can be broadly categorized into two types. One type prioritizes nodes based on their graph structure attributes, such as degree and betweenness centrality, selecting the optimal node as the SR node. This method is relatively simple but ineffective because it only considers topology without taking into account actual traffic flow. The other type selects a suitable intermediate node for each flow in the traffic set, upgrading it to an SR node with a limited number of upgradable nodes, aiming to minimize maximum link utilization. This optimization problem is typically modeled and solved using linear programming. However, this flow-by-flow SR node selection method ignores learnable information such as network topology and the constraints between SR node distributions, merely choosing the best combination from multiple possibilities, thus limiting it to linear programming. For scenarios where traffic can pass through multiple intermediate SR nodes, linear programming results in a huge problem space and complex computations. Therefore, this method typically restricts a single flow to passing through at most one intermediate SR node, i.e., two routes (2-SR). Clearly, this has significant limitations in large topologies containing many long-path traffic. Summary of the Invention

[0005] To address the problems existing in the above-mentioned methods for partially deploying SR-TE, namely that traffic path planning only considers the network topology without considering the actual traffic situation, or restricts each traffic flow to pass through two routes while ignoring the network topology, this invention proposes a method for partially deploying SRv6 nodes and traffic planning based on celestial balance, which realizes traffic path planning based on celestial balance theory.

[0006] This invention proposes a method for partial deployment of SRv6 nodes and flow planning based on celestial equilibrium. It analogizes the influence of SRv6 nodes to the gravitational pull of a star on surrounding planets in celestial equilibrium theory, thus modeling the impact of SRv6 nodes on surrounding ordinary nodes. The method delineates the influence region of SRv6 nodes based on their influence, analogous to how a star attracts surrounding planets to condense and form a galaxy through gravity. Specifically, the method selects SRv6 nodes for partial deployment based on their influence, and the implementation steps include:

[0007] Step 1: Analogize the influence of SRv6 nodes to the gravitational pull of stars on surrounding planets in celestial equilibrium theory. Based on the influence of SRv6 nodes, select some SRv6 nodes for partial deployment, including the following sub-steps:

[0008] (1) Initially, n network nodes are randomly selected in the network topology as SRv6 nodes, and the value of the influence factor k of each SRv6 node is randomly initialized; where n is a positive integer and the value of k is in the range of [1,9].

[0009] (2) Calculate the influence of each SRv6 node on ordinary nodes that are not SRv6 nodes in the network topology, and divide the ordinary nodes into the influence area of ​​the SRv6 node that has the greatest influence; calculate the influence of the SRv6 node on node a F=k*load / r, where load represents the average link utilization of the link directly connected to the SRv6 node, and r represents the distance between the SRv6 node and node a.

[0010] (3) Perform step two to calculate the path of each traffic flow, and then recalculate the load of all links in the network topology;

[0011] (4) Perform iterative E-step: Based on the current set of influencing factors, reselect SRv6 nodes for each influencing region. The selection strategy is to assume each node in the influencing region as an SRv6 node, calculate the link load, and select the node that minimizes the link load as the SRv6 node for that region.

[0012] (5) Perform M-step iteration: Based on the current set of SRv6 nodes, reselect the value of the influence factor k for each SRv6 node. The selection strategy is to change the value of the influence factor k for each SRv6 node, calculate the link load, and select the influence factor that minimizes the link load as the optimal influence factor for that SRv6 node.

[0013] (6) Repeat the E-step and M-step iterations until the distribution of SRv6 nodes and their affected regions tends to converge, then end the iteration and output the final set of SRv6 nodes and the set of influence factors.

[0014] Step 2: Based on the celestial equilibrium theory, perform traffic planning and calculate the path for each traffic flow. This includes: for each traffic flow in the current traffic set, calculate the sequence of influence areas traversed by the shortest path, and select the SRv6 nodes corresponding to the sequence of influence areas as candidate relay nodes for that traffic flow; from the candidate relay node set for each traffic flow, select the set of SRv6 nodes that provides the best network load balancing effect as the actual relay node set for that traffic flow; and construct the routing path for each traffic flow based on the actual relay node set for each traffic flow.

[0015] In step two, the load of all links in the current network topology is obtained. Then, for each traffic in the traffic set, a greedy algorithm is used to select the set of SRv6 nodes that provides the best network load balancing effect. The greedy condition is: if the traffic passes through a candidate relay node and reduces the maximum link load in the network topology, then the candidate relay node is added to the actual relay node set for that traffic. When constructing the routing path for each traffic, the shortest path method is used to determine the route between two adjacent actual relay nodes.

[0016] The advantages of this invention are as follows: Existing node-based graph structure attribute sorting methods are relatively simple, but their effectiveness is poor, as they do not consider actual traffic conditions. Another type of optimization method based on linear programming generates a huge problem space and complex computational load; therefore, this method usually limits traffic to passing through at most one SR node, failing to fully utilize the traffic splitting capacity of SRv6 nodes. Compared with existing technologies, this invention introduces celestial equilibrium theory from a global perspective, combining global topology, traffic distribution, and other information to optimize the distribution of SRv6 nodes. Specifically, it analogizes the influence of SRv6 nodes on surrounding ordinary nodes to gravity in celestial equilibrium theory, forming an influence region for each SRv6 node, and iteratively optimizes the distribution of SRv6 nodes and their influence regions to achieve optimal network load balancing. This invention combines various information such as global network topology and traffic distribution, and does not limit each traffic to only two routes; each traffic can be relayed to multiple intermediate SRv6 nodes, achieving a superior network load balancing effect. Attached Figure Description

[0017] Figure 1 This is an example diagram illustrating the calculation of SRv6 influence in an embodiment of the present invention;

[0018] Figure 2 This is an example diagram of the SRv6 influence area defined in an embodiment of the present invention;

[0019] Figure 3 This is a flowchart of the method for calculating network load in this invention;

[0020] Figure 4 This is a flowchart of the E-step and M-step iterations of the method of the present invention;

[0021] Figure 5 This is a flowchart of the flow planning process in the method of this invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0023] This invention proposes a method for partial deployment SRv6 node selection and traffic planning based on celestial equilibrium theory. The influence of SRv6 nodes is analogous to the gravitational pull of stars on surrounding planets in a galaxy, modeling the impact of SRv6 nodes on surrounding ordinary nodes. The influence of SRv6 nodes on other ordinary nodes includes multiple factors: the load of links surrounding the SRv6 node, the structural attributes of the SRv6 node in the network topology, and the distance between the SRv6 node and ordinary nodes. The formula for calculating the influence F of an SRv6 node on a particular ordinary node is: F = k * load / r, where k represents the influence factor of the SRv6 node; a larger value indicates a larger influence range. The value of k is learned through an algorithm. Load represents the average link utilization of the links directly connected to the SRv6 node, and r represents the distance between the SRv6 node and ordinary nodes.

[0024] This invention divides the influence regions of SRv6 nodes by their influence, analogous to the celestial equilibrium theory where stars attract surrounding planets to form a galaxy through gravity. In the network topology, SRv6 nodes attract surrounding ordinary nodes to form their own influence regions, thus modeling the range of influence of SRv6 nodes on ordinary nodes. The influence of each SRv6 node on all other ordinary nodes is calculated according to the influence calculation formula. If an ordinary node is most influenced by one particular SRv6 node, then that node belongs to that SRv6 node's influence region. Once all ordinary nodes have been divided, the influence regions of all SRv6 nodes in the network topology are simultaneously formed. For example... Figure 1 As shown, for example, nodes numbered 5, 7, 9, and 14 are selected as SRv6 nodes, and the influence of each SRv6 node on other ordinary nodes, such as node 4, is calculated. Figure 2 As shown, the influence area of ​​SRv6 nodes is divided according to the calculated influence. Different colors in the figure represent different areas. For example, node 4 is divided into the influence area of ​​node 5.

[0025] This invention implements a partial deployment SRv6 node selection method based on celestial equilibrium. It uses an iterative algorithm to select the SRv6 node with the optimal load balancing effect and its influence range, comprising an E-step and an M-step: The E-step, combined with the influence range obtained from the M-step, reselects a new SRv6 node with the optimal load balancing effect within the influence region, obtaining a new influence region; The M-step, combined with the SRv6 node set obtained from the E-step, reselects the optimal influence factor for load balancing effect for each SRv6 node. The E-step and M-step iterative steps are repeated, analogous to the celestial equilibrium process, until the distribution of SRv6 nodes and their influence regions converges, achieving the optimal network load balancing effect. This invention implements a traffic path planning method based on the theory of celestial equilibrium. It sequentially processes each traffic flow in the traffic set according to the sequence of SRv6 node influence regions traversed by the shortest path. Analogous to an alien object being attracted into a galaxy, a traffic flow passing through the influence region of a certain SRv6 node has a high probability of being redirected to that SRv6 node for traffic distribution; therefore, that SRv6 node is considered a candidate redirection node for this traffic. Based on the obtained sequence of SRv6 node influence regions traversed by each traffic flow, the set of SRv6 nodes that provide the best network load balancing effect for each traffic flow is selected, thereby constructing the routing path for each traffic flow.

[0026] like Figure 3 and Figure 4 As shown, the partial deployment SRv6 node selection method based on celestial balance implemented in this embodiment of the invention includes the following steps 1-8.

[0027] Step 1) Randomly select n SRv6 nodes in the network topology and randomly initialize the value of the influence factor k of each SRv6 node, where n is the maximum number of nodes that can be upgraded to SRv6, and its value range can be from 1 to the number of all nodes in the network topology. The specific value of n can be set according to actual requirements.

[0028] Step 2) Calculate the influence of each SRv6 node on all ordinary nodes in the network topology, F=k*load / r, where k represents the influence factor of the SRv6 node; load represents the average link utilization of the directly connected links of the SRv6 node, that is, the ratio of the sum of the loads of all directly connected links of the SRv6 node to the number of directly connected links, where the link load is the ratio of the current traffic volume on the link to the link capacity, expressed as LU=load(e) / C(e), where load(e) is the traffic volume on link e, and C(e) is the capacity of link e; r represents the distance between the SRv6 node and the ordinary node.

[0029] Step 3) Based on the SRv6 node corresponding to the largest influence on each ordinary node, divide the ordinary node into the influence area of ​​that SRv6 node, so that n influence areas of SRv6 nodes are formed in the network topology.

[0030] Step 4) Calculate the path of each flow using the flow planning method based on celestial equilibrium theory proposed in this invention.

[0031] Step 5) Based on the routing paths of all traffic obtained in the previous step, recalculate the load LU on all links in the network topology. Use the maximum link load to measure the load balancing effect. The smaller the maximum link load, the better the network load balancing effect.

[0032] Step 6) Iteration E-step: Based on the set of influence factors obtained from iteration M-step, and combined with the network load calculation methods defined in steps 2 to 5, reselect the set of SRv6 nodes with the best load balancing optimization effect in each of the n influence regions. Within each influence region, for each node, assume it as an SRv6 node and calculate the maximum link load value LU in the corresponding global network topology. Select the node that minimizes the maximum link load value LU in the global network topology as the SRv6 node for that region to achieve load balancing optimization. If this is the first iteration, the set of influence factors is randomly initialized.

[0033] In this embodiment of the invention, during the E-step iteration, a new SRv6 node is selected for each affected region, that is, the position of the SRv6 node in each affected region is changed. After each change of the SRv6 node in the affected region, steps 2-5 need to be executed again.

[0034] Step 7) Iteration M step: Combining the SRv6 node set obtained from iteration E step, and according to the network load calculation method defined in steps 2 to 5, reselect the influence factor that can optimize the load balancing effect for each SRv6 node: For each SRv6 node, assume that its influence factor value ranges from 1 to 9, and calculate the maximum link load value LU in the corresponding global network topology. Select the influence factor that minimizes the maximum link load value LU in the global network topology as the optimal influence factor for the SRv6 node, thereby optimizing the load balancing effect.

[0035] In this embodiment of the invention, the value range of the influence factor k is [1, 9], and positive integers can be used for simplification. During the M-step iteration, the value of the influence factor is changed for each SRv6 node. Each time the value of the influence factor is changed, steps 2-5 above must be re-executed.

[0036] Step 8) Repeat the E-step and M-step iterations, analogous to the celestial equilibrium process, until the distribution of SRv6 nodes and their affected areas tends to converge, achieving the optimal network load balancing effect. Output the final set of SRv6 nodes and the set of influencing factors, and use the traffic planning method based on celestial equilibrium theory to calculate the path of each traffic flow.

[0037] This invention designs and implements a flow planning method based on celestial equilibrium theory, such as... Figure 5 As shown, steps 41-44 are included.

[0038] Step 41) Calculate the sequence of SRv6 node influence areas that each traffic flow passes through in the traffic set according to the shortest path; each traffic flow is a data stream transmitted from a source node to a destination node.

[0039] Step 42) Analogous to an alien object being attracted into a galaxy, if a flow passes through the influence area of ​​a certain SRv6 node, then this flow is very likely to be diverted to that SRv6 node to achieve a diversion effect. Therefore, this SRv6 node will be used as a candidate diversion node for this flow. Based on the sequence of influence areas of each flow through the SRv6 node obtained in the previous step, the set of candidate diversion nodes for this flow is obtained.

[0040] Step 43) Select the SRv6 node set that has the best effect on network load balancing from the candidate relay node set for each traffic, and use it as the actual relay node set for this traffic: Use a greedy algorithm to select. If a traffic passes through a certain candidate relay node and makes the network load more balanced, that is, reduces the maximum link load LU among all links in the network topology, then this candidate relay node is added to its actual relay node set by this traffic.

[0041] In this embodiment of the invention, for each traffic flow, a greedy algorithm is used to offline select the set of SRv6 nodes that provides the best network load balancing effect. The greedy condition is: if selecting a candidate relay node as a route can reduce the maximum load in the network, then that node is selected. The actual relay node set is then selected sequentially for each traffic flow in the traffic set.

[0042] Step 44) Construct the routing path for each traffic based on the actual relay node set for each traffic: In the actual relay node set, traffic is routed using the shortest path between two adjacent actual relay nodes.

[0043] like Figure 2As shown, a traffic flow has a source node of 1 and a destination node of 6, with its shortest path being from node 1 to node 2 to node 6. According to the traffic planning method based on celestial equilibrium theory proposed in this invention, this traffic flows through the influence areas of SRv6 nodes 7 and 5 sequentially along the shortest path. Therefore, SRv6 nodes 7 and 5 become candidate relay nodes for this traffic. A greedy algorithm is used to select the actual SRv6 node that this traffic flows through. Assuming that the traffic does not make the network load more balanced when passing through SRv6 node 7, but does make the network load more balanced when passing through SRv6 node 5, then the actual SRv6 node that this traffic flows through is node 5. Following the shortest path routing between two adjacent actual relay nodes, the path of this traffic is from node 1 to node 4 to node 5, and then from node 5 to the destination node 6.

[0044] Except for the technical features described in the specification, all other technologies are known to those skilled in the art. Descriptions of well-known components and technologies are omitted in this invention to avoid redundancy and unnecessary limitation. The embodiments described above do not represent all embodiments consistent with this application. Various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this invention are still within the protection scope of this invention.

Claims

1. A method for partial deployment of SRv6 nodes and traffic planning based on celestial equilibrium, comprising: Step 1: Analogize the influence of SRv6 nodes to the gravitational pull of stars on surrounding planets in celestial equilibrium theory. Based on the influence of SRv6 nodes, select some SRv6 nodes for partial deployment, including the following sub-steps: (1) Initially, n network nodes are randomly selected in the network topology as SRv6 nodes, and the value of the influence factor k of each SRv6 node is randomly initialized; where n is a positive integer and the value of k is in the range of [1,9]. (2) Calculate the influence of each SRv6 node on the ordinary nodes that are not SRv6 nodes in the network topology, and divide the ordinary nodes into the influence regions of the SRv6 nodes that are most affected. The influence of an SRv6 node on a normal node is calculated as F = k * load / r, where load represents the average link utilization of the links directly connected to the SRv6 node, and r represents the distance between the SRv6 node and the normal node. (3) Perform step two to calculate the path of each traffic flow, and then recalculate the load of all links in the network topology; (4) Perform iterative E-step: Based on the current set of influencing factors, reselect SRv6 nodes for each influencing region. The selection strategy is to assume each node in the influencing region as an SRv6 node, calculate the link load, and select the node that minimizes the link load as the SRv6 node for that region. (5) Perform M-step iteration: Based on the current set of SRv6 nodes, reselect the value of the influence factor k for each SRv6 node. The selection strategy is to change the value of the influence factor k for each SRv6 node, calculate the link load, and select the influence factor that minimizes the link load as the optimal influence factor for that SRv6 node. (6) Repeat the E-step and M-step iterations until the distribution of SRv6 nodes and their affected regions tends to converge, then end the iteration and output the final set of SRv6 nodes and the set of influence factors. Step 2: Based on the celestial equilibrium theory, perform traffic planning and calculate the path for each traffic flow. This includes: for each traffic flow in the current traffic set, calculate the sequence of influence areas traversed by the shortest path, and select the SRv6 nodes corresponding to the sequence of influence areas as candidate relay nodes for that traffic flow; from the candidate relay node set for each traffic flow, select the set of SRv6 nodes that provides the best network load balancing effect as the actual relay node set for that traffic flow; and construct the routing path for each traffic flow based on the actual relay node set for each traffic flow.

2. The method of claim 1, wherein, In step one, the average link utilization load of the SRv6 node is calculated as the ratio of the sum of the loads of the links directly connected to the SRv6 node to the number of the directly connected links. The load on a link is calculated as the ratio of the traffic volume on that link to the link's capacity.

3. The method of claim 1, wherein, In step one, during the execution of iteration E, an SRv6 node is reselected for each affected region, and then sub-steps 2 and 3 are re-executed to obtain the link load.

4. The method according to claim 1 or 3, characterized in that, In step one, during the M-step iteration, the value of the influence factor k is changed for each SRv6 node, and then sub-steps 2 and 3 are re-executed to obtain the link load.

5. The method according to claim 1, characterized in that, In step two, the load of all links in the current network topology is obtained. Then, for each traffic in the traffic set, a greedy algorithm is used to select the SRv6 node set with the best network load balancing effect. The greedy condition is: if the traffic passes through a candidate relay node, the maximum link load in the network topology is reduced, then the candidate relay node is added to the actual relay node set of the traffic.

6. The method according to claim 1, characterized in that, In step two, when constructing the routing path for each traffic flow, the shortest path method is used to determine the route between two adjacent actual relay nodes.

Images