Comprehensive path optimization method for wide-area cloud architecture network

Abstract

The invention relates to a comprehensive path optimization method for a wide-area cloud architecture network in the field of network cloud fusion. The method is particularly applicable to load balancing scheduling of a large number of continuous service flows (elephant flows) in a cloud architecture network. Comprising the following steps: monitoring a network state and identifying elephant flow, collecting information of each link, and detecting and identifying the elephant flow; topological path pre-screening and candidate path library construction improve the calculation and optimization efficiency. Initializing ant colony parameters, and formulating a multi-dimensional heuristic function; performing path exploration and probability selection, and generating an ant colony exploration calculation formula according to a target; dynamically updating pheromones, and searching an optimal path according to local pheromone updating and global pheromone updating; issuing an optimal path and updating a flow table; and continuous monitoring and dynamic adjustment are carried out. According to the invention, a dual control plane combining a routing protocol and an SDN protocol is designed, the improved ant colony algorithm based on network topology is adopted to optimize elephant flow path selection under a cloud network architecture, and the network global service transmission efficiency is improved.

Description

TECHNICAL FIELD

[0001] The application relates to a comprehensive path optimization method for a wide-area cloud architecture network. BACKGROUND

[0002] With the gradual maturity of application scenarios such as data centers, cloud computing and mobile Internet, wide-area networks develop into cloud architecture networks, and traditional single control plane path selection has limitations, such as poor traffic scheduling flexibility and low link resource utilization in a routing protocol control plane, and problems such as too large global information collection and low center fault tolerance in an SDN control plane. Especially when elephant flows (large and continuous) and mouse flows (small and short) appear in the same network, the transmission quality of key services in the mouse flows cannot be guaranteed. Therefore, it is particularly important to design a double control plane combining a routing protocol and an SDN protocol, and to optimize elephant flow path selection in a cloud architecture network based on a network topology improved ant colony algorithm, so as to improve the global service transmission efficiency. SUMMARY

[0003] The application proposes a comprehensive path optimization method for a wide-area cloud architecture network to solve the problem that elephant flows in a double control plane affect the transmission quality of key services, realizes global scheduling for elephant flows, and improves network resource availability.

[0004] The technical scheme adopted by the application is as follows:

[0005] A comprehensive path optimization method for a wide-area cloud architecture network comprises the following steps:

[0006] Step 1: Each switch collects and summarizes the utilization rate, delay and packet loss rate of each link in real time to an SDN controller. When it is monitored that the bandwidth occupation of a certain data flow continuously exceeds a set threshold value and the total data volume exceeds a set value, the data flow is marked as an elephant flow, and step 2 is executed.

[0007] Step 2: The SDN controller calculates the path criticality according to the link type and bandwidth design weight, sorts the links, calculates the link utilization rate, removes the links with a utilization rate higher than a threshold value, and selects a set number of links from the remaining links according to the sorting to form an initial path set.

[0008] Step 3: The ant colony parameters are initialized, including the ant colony size, initial value of pheromone, pheromone evaporation factor, pheromone factor, heuristic factor and iteration number, and a multi-dimensional heuristic function is constructed.

[0009] Step 4, according to the multi-dimensional heuristic function and pheromone, the transition probability is calculated, each ant starts from the source node, and selects the next hop node in the initial path set according to the transition probability;

[0010] Step 5, after each ant moves to the next hop node, the local pheromone is updated, and then returns to step 4 until it moves to the terminal node;

[0011] Step 6, after all ants move to the terminal node, the global pheromone is updated according to the fitness of the path, and then returns to step 4 until the iteration number is reached, and the final path is obtained;

[0012] Step 7, the SDN controller converts the optimal path obtained after iteration convergence into OpenFlow flow table entries and issues them to the relevant switches to execute the elephant flow rerouting.

[0013] Further, in step 2, the path criticality calculation method is:

[0014]

[0015] In the formula, is the microwave link weight, is the satellite link weight, is the wired link weight, is the number of microwave links on the path, is the number of satellite links on the path, is the number of wired links on the path;

[0016] The link utilization rate calculation method is:

[0017] C = max(L a / L max )

[0018] In the formula, L a is the real-time load of the link in the path p, and L max is the link capacity.

[0019] Further, in step 3, the multi-dimensional heuristic function is:

[0020]

[0021] In the formula, a, b and c are weights, is the remaining bandwidth of nodes i and j in the link, is the total bandwidth of nodes i and j in the link, D ij is the delay of nodes i and j in the link, is the maximum delay of nodes i and j in the link, LS ij is the packet loss rate of nodes i and j in the link.

[0022] Further, in step 4, the transition probability is:

[0023]

[0024] where α is pheromone factor, β is heuristic factor, is the pheromone of link between node i and node j, allowed k is all links in initial path set.

[0025] Further, in step 5, the local pheromone update method is:

[0026]

[0027] where ρ is pheromone evaporation factor, Δτ is pheromone increment, and are pheromone before and after update respectively.

[0028] Further, in step 6, the global pheromone update method is:

[0029]

[0030]

[0031]

[0032] where ρ is pheromone evaporation factor, Q is constant, is the total length of path constructed by ant k this time, m is the size of ant colony, and are pheromone before and after update respectively.

[0033] Compared with the background art, the present application has the following advantages:

[0034] 1) Deep integration and closed-loop control of SDN and ant colony algorithm: the global network view and centralized control capability of SDN are used to provide real-time and comprehensive network state data for the ant colony algorithm, which is directly and quickly applied to the network through the flow table.

[0035] 2) Differentiated scheduling for elephant flow: the elephant flow which has a significant impact on network performance is centrally optimized and rerouted, which can more effectively avoid network congestion and ensure the transmission quality of critical services compared with the traditional routing method.

[0036] 3) Reduce search space to improve convergence speed: the reduction of the number of optional paths through the known network link quality (type, bandwidth, etc.) directly reduces the calculation amount of state transition probability and the decision time of ants constructing paths

[0037] 4) Load-aware heuristic function: The network state information such as bandwidth utilization, latency, packet loss rate of the link is integrated into the heuristic function of the ant colony algorithm in a weighted manner to establish a multi-dimensional link evaluation model, so that the path selection can adapt to the network traffic changes, and not just rely on static distance. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a flowchart of the present application. DETAILED DESCRIPTION

[0039] The technical solutions of the present application will be further described in detail below with reference to the accompanying drawings.

[0040] Referring to Figure 1 , a comprehensive path optimization method for a wide-area cloud architecture network, the specific steps include:

[0041] 1) Network state monitoring and elephant flow identification. After the initialization of the SDN controller is completed, each switch collects and summarizes the utilization rate, latency and packet loss rate of each link to the SDN controller in real time, and when a new data flow is monitored, the average bandwidth occupancy flow=(byte2-byte1) / (t2-t1) / BW is calculated within the time window Δt, wherein BW is the maximum bandwidth. If flow exceeds the preset value such as 10% of the total link, and the total data volume is large, it is marked as an elephant flow, and the subsequent scheduling process is triggered. If there is no elephant flow, the default path is taken.

[0042] 2) Topology path pre-screening and candidate path library construction. Combined with the link state collection information of the switch in step 1), the weights of the link types and bandwidth are designed, for example: the weight of wired is higher than that of microwave, and the weight of microwave is higher than that of satellite communication. The k paths are optimized by calculating the path criticality.

[0043]

[0044] In the formula, is the weight of the microwave link, is the weight of the satellite communication link, is the weight of the wired link, is the number of microwave links on the path, is the number of satellite communication links on the path, is the number of wired links on the path;

[0045] Then the link utilization rate C=max(L a / L max ) is calculated, wherein L a is the real-time load of the link in the path p, and L maxFor link capacity, paths with utilization higher than a threshold (e.g. 0.8) are removed to avoid selecting potentially congested paths. Through traversing 3-5 paths in the remaining links according to the ranking, a path search space for subsequent ant colony algorithm is constructed.

[0046] 3), initialize ant colony parameters and heuristic function. The ant colony parameters include ant colony size m, pheromone initial value, pheromone evaporation factor p, pheromone factor (importance a), heuristic factor b and iteration number; wherein the initial value of path pheromone in the search space is set to 1, and the non-candidate path is set to 0.01, guiding ants to focus on the search space.

[0047] Elephant flow focuses on bandwidth, sets bandwidth weight a> delay weight b> packet loss rate weight c (e.g. b=0.6, d=0.3, l=0.1), W is bandwidth, D is delay, LS is packet loss rate; construct multi-dimensional heuristic function:

[0048] ;

[0049] In the formula, a, b and c are weights, is the remaining bandwidth of node i and node j in the link, is the total bandwidth of node i and node j in the link, D ij is the delay of node i and node j in the link, is the maximum delay of node i and node j in the link, LS ij is the packet loss rate of node i and node j in the link.

[0050] 4), path exploration and probability selection. According to the multi-dimensional heuristic function and pheromone, the transition probability is calculated, and each ant starts from the source node and selects the next hop node in the initial path set according to the transition probability;

[0051] The transition probability is:

[0052] ;

[0053] In the formula, is the pheromone of node i and node j in the link, allowed k is all links in the initial path set, a is the pheromone factor (usually 1), b is the heuristic factor (2-3 in the elephant flow scenario), elite ant strategy can be selected, 20% of the ants are selected as elite ants, only explore on the current optimal path in the iteration, and strengthen positive feedback; the rest of the ants randomly explore in the candidate path library to maintain diversity.

[0054] 5), dynamic update of pheromone. After each ant moves to the next hop node, local pheromone update is performed, and then step 4) is returned until it moves to the terminal node;

[0055] After each passing of the link (i, j), the ants immediately update the pheromone as:

[0056]

[0057] Update the pheromone, where p is the evaporation coefficient (0.3-0.6), and Δτ is the pheromone increment (e.g. 0.1), and are the pheromone before and after the update, respectively.

[0058] Global update. After all ants move to the terminal node, the global pheromone update is performed according to the fitness of the path, and then return to step 4) until the number of iterations is reached, and the final path is obtained;

[0059] After the local iteration is completed, only the optimal path and the elite path in the candidate path library are enhanced:

[0060]

[0061]

[0062]

[0063] where the new pheromone concentration uses the ant cycle model, Q is a constant, is the total length of the path constructed by ant k this time, and m is the size of the ant colony;

[0064] When the optimal path does not change for 3 consecutive iterations, or the total number of iterations reaches 50, the algorithm is determined to be converged.

[0065] 6), optimal path distribution and flow table update: the SDN controller converts the final optimal path into specific flow table entries through OpenFlow and other protocols, and distributes them to the relevant switches on the path. The flow table entry matches the five-tuple information (source IP, destination IP, source port, destination port, protocol type) of the elephant flow, and specifies the corresponding egress port.

[0066] 7), continuous monitoring and dynamic adjustment. If it is found that the utilization rate of a certain link continuously exceeds 85%, an alarm is triggered. If the link exists in the current candidate path library, the path is marked as "congested path" and temporarily removed from the search space, and a new round of scheduling process is immediately started to calculate a new path for the affected flow, and step 4) is entered.

Claims

1. A comprehensive path optimization method for wide-area cloud-based network architectures, characterized in that, Includes the following steps: Step 1: Each switch collects and summarizes the utilization, latency and packet loss rate of each link in real time to the SDN controller. When it is detected that the bandwidth usage of a certain data flow continuously exceeds the set threshold and the total data volume exceeds the set value, it is marked as an elephant flow and Step 2 is executed. Step 2: The SDN controller designs weights based on link type and bandwidth, calculates path criticality, sorts each link, calculates link utilization, removes links with utilization exceeding a threshold, and selects a set of links from the remaining links according to the sorting to form an initial path set. Step 3: Initialize ant colony parameters, including ant colony size, initial pheromone value, pheromone evaporation factor, pheromone factor, heuristic factor, and number of iterations, and construct a multi-dimensional heuristic function; Step 4: Calculate the transition probability based on the multi-dimensional heuristic function and pheromone. Each ant starts from the source node and selects the next hop node from the initial path set according to the transition probability. Step 5: Each time the ant moves to the next hop node, it performs a local pheromone update and then returns to step 4 until it moves to the final node. Step 6: After all ants have moved to the terminal node, perform a global pheromone update based on the fitness of the path, and then return to step 4 until the number of iterations is reached to obtain the final path. Step 7: The SDN controller converts the optimal path obtained after iterative convergence into an OpenFlow flow table entry and sends it to the relevant switches to perform rerouting of the elephant flow.

2. The comprehensive path optimization method for wide-area cloud architecture networks according to claim 1, characterized in that, In step 2, the path criticality is calculated as follows: In the formula, For microwave link weights, For satellite communication link weight, For wired link weights, This represents the number of microwave links on the path. This represents the number of satellite communication links along the path. This represents the number of wired links on the path. The link utilization rate is calculated as follows: C=max(L a / L max ) In the formula, L a For the real-time load of the links in path p, L max This refers to the link capacity.

3. The comprehensive path optimization method for wide-area cloud architecture networks according to claim 1, characterized in that, In step 3, the multi-dimensional heuristic function is: In the formula, a, b, and c are weights. Let be the remaining bandwidth of nodes i and j in the link. D represents the total bandwidth of nodes i and j in the link. ij Let be the time delay between node i and node j in the link. Let LS be the maximum delay between node i and node j in the link. ij Let be the packet loss rate of nodes i and j in the link.

4. The comprehensive path optimization method for wide-area cloud architecture networks according to claim 1, characterized in that, In step 4, the transition probability is: In the formula, α is the pheromone factor, and β is the heuristic factor. For the pheromones of nodes i and j in the link, allowed k This refers to all links in the initial path set.

5. The comprehensive path optimization method for wide-area cloud architecture networks according to claim 1, characterized in that, In step 5, the local pheromone update method is as follows: In the formula, ρ is the pheromone evaporation factor, and Δτ is the pheromone increment. and These are the pheromones before and after the update, respectively.

6. The comprehensive path optimization method for wide-area cloud architecture networks according to claim 1, characterized in that, In step 6, the global pheromone update method is as follows: In the formula, ρ is the pheromone evaporation factor, and Q is a constant. Here, m is the total length of the path constructed by ant k in this operation, and m is the size of the ant colony. and These are the pheromones before and after the update, respectively.

Images