An efficient backup method for service chain based on cloud environment

Abstract

The present invention relates to an efficient backup method for a service chain based on a cloud environment, and belongs to the technical field of CSC efficient backup. The method first models the CSC reliability backup problem as a device site selection model, and designs an efficient redundant deployment algorithm based on a taboo search strategy to achieve the reduction of backup virtual nodes. Then, the mapping relationship between the primary and standby VNF instances is optimized through a joint game algorithm to reduce the synchronization link delay of the primary and standby VNF instances. The use example proves that the backup method can effectively reduce operating expenses and improve the synchronization delay between the primary and standby virtual network function instances.

Description

Technical Field

[0001] The present invention belongs to the technical field of CSC efficient backup, and in particular relates to a service chain efficient backup method based on a cloud environment. Background Art

[0002] Cloud environments can use virtualization technology to decouple network functions from dedicated hardware and run them in a virtual resource pool in the form of virtual network functions (VNFs), replacing the closed, dedicated, and rigid hardware devices in the current network. This allows network functions to share underlying resources, be deployed in real time according to business needs, elastically scale, self-heal, and isolate risks, and achieve rapid service provision and fine-grained resource management and control. In a cloud environment, building a cloud service chain to provide network services greatly improves the flexibility and scalability of network service deployment, bringing opportunities for the innovation of traditional network services. A path formed by a group of virtual network functions connected in series according to specific service requirements is called a cloud service chain (CSC). Figure 1 The following shows a CSC instance in a cloud environment. The CSC consists of a firewall, an IDS, and a proxy. The data initiated by the user needs to be processed by the above network function components in sequence to achieve network access.

[0003] However, compared with highly robust dedicated hardware devices, the resource sharing characteristics of the cloud environment bring certain survivability risks to CSC. In the case of software and hardware failures and malicious attacks, the frequency of single node or single link failures in the underlying network is relatively high. Although link failures can be quickly recovered by adjusting the traffic path strategy, node failures involve the failure of bearer resources, which brings more fatal damage to the chain-structured CSC. Damage to any network function node can cause service termination. To ensure the survivability of CSC, designing fault recovery and fault tolerance strategies is an inevitable choice.

[0004] At present, the research work on SFC survivability can be summarized into two types: fault recovery and redundant backup. Among them, the fault recovery strategy does not need to pre-place spare resources in the underlying network. When a fault occurs, the fault recovery strategy implements fault elimination and service recovery by deploying spare instances or updating traffic steering strategies. The redundant backup strategy requires pre-placement of spare VNF instances in the underlying network, and real-time synchronization between the primary and standby instances. When a fault occurs, the standby instance can quickly switch from the standby state to the running state synchronized with the primary instance to ensure service continuity. Although the fault recovery strategy can save spare resources, it will face non-negligible deployment delays and state synchronization problems when a fault occurs, which will also cause long-term service interruptions. Therefore, operators are more inclined to adopt redundant backup strategies to provide better SFC survivability.

[0005] Existing solutions for CSC survivability can be summarized into two types: fault recovery and redundant backup. The fault recovery strategy refers to the implementation of CSC recovery by deploying a recovery image or updating the traffic path strategy when a fault occurs. The redundant backup strategy requires pre-setting backup VNF instances in different node resource pools, and maintaining phased synchronization between the primary and backup VNF instances. When a fault occurs, the backup VNF instance can quickly switch to a running state synchronized with the primary VNF instance to ensure rapid service recovery. In comparison between these two methods, the fault recovery strategy can save backup resources, but will face non-negligible deployment delays and state synchronization issues, and the service interruption time cannot be ignored. Therefore, operation and maintenance personnel prefer redundant backup strategies to ensure better CSC survivability.

[0006] The existing research work on CSC backup strategies has achieved certain results, but there are still deficiencies in dealing with the problem of underlying node failures. The literature "Carella G. Resilient Orchestration of Service Functions Chains in a NFV Environment [C] / / Network Function Virtualization & Software Defined Networks. IEEE, 2017" restores the failed CSC by enabling backup virtual network function instances and updating traffic paths, but no specific plan is planned. The design scheme of the literature "Suh D, Baek H, Jang S, et al. Distributed Service Function Failover Mechanism in Service Function Chaining [C] / / International Conference on Information Networking. 2017" deploys the backup virtual network function instance in the server connected to the same switch. This scheme ensures the synchronization of the active and standby VNFs to achieve fast fault recovery, but the situation where there is no random node failure is not taken into consideration. The paper "Fan J, Gao X, Ye Z, et al. GREP: Guaranteeing Reliability with Enhanced Protection in NFV[C] / / Acm Sigcomm Workshop on Hot Topics in Middleboxes & Network Function Virtualization. ACM, 2015" measures the survivability of CSC by the probability that all virtual network functions in CSC will not fail when a node fails, and proposes a mapping algorithm based on joint backup VNF, but it cannot cope with the failure of underlying network nodes in cloud environments. Since any node in the underlying network may fail, the redundant backup strategy inevitably pre-sets backup VNF instances to ensure the survivability of CSC, so it is necessary to design an algorithm to optimize the overhead caused by backup resources.The paper "Yu H, Anand V, Qiao C, et al. Cost Efficient Design of Survivable Virtual Infrastructure to Recover from Facility Node Failures[C] / / Proceedings of IEEE International Conference on Communications, ICC 2011, Kyoto, Japan, 5-9 June, 2011. IEEE, 2011" proposed a backup solution for the problem of single node failure of underlying resources in cloud environment. The design was based on swarm intelligence heuristic algorithm to solve the optimal number of backups, but did not take the state synchronization problem between the active and standby VNFs into consideration. Summary of the invention

[0007] 1. Technical issues to be resolved

[0008] The present invention aims to solve the problem of excessive resource overhead in a survivable CSC redundant backup strategy in a cloud environment and proposes an efficient backup method for a service chain.

[0009] (II) Technical solution

[0010] In order to solve the above technical problems, the present invention provides a service chain efficient backup method based on a cloud environment, comprising the following steps:

[0011] Step 1: Establish a mathematical model for the cloud environment elements and decision variables that constitute the service chain in the cloud environment, and determine the constraints and optimization goals of the cloud service chain CSC backup method;

[0012] Step 2: Based on step 1, the backup deployment algorithm based on the capacity-constrained equipment site selection CFLP model is used to abstract the survivability problem of CSC into a site selection problem, and the backup redundancy is compressed based on the taboo search algorithm.

[0013] Step 3: Based on step 2, a joint game strategy is used to optimize the association between the active and standby VNF instances to reduce the synchronization link delay between the active and standby VNF instances.

[0014] The invention also provides a system for implementing the method.

[0015] The present invention also provides a performance evaluation method, which is used to evaluate the performance of the method.

[0016] (III) Beneficial effects

[0017] In view of the problem of excessive resource overhead in the redundant backup strategy of survivable CSC in cloud environment, the present invention proposes an efficient backup method for service chain based on cloud environment. The method first models the CSC reliability backup problem as a device site selection model, and designs an efficient redundant deployment algorithm based on the taboo search strategy to achieve the reduction of backup virtual nodes. Then, the mapping relationship between the primary and standby VNF instances is optimized through the joint game algorithm to reduce the synchronization link delay of the primary and standby VNF instances. The use example proves that the backup (sharing) method can effectively reduce operating expenses and improve the synchronization delay between the primary and standby virtual network function instances. The method can effectively reduce operating expenses and optimize the synchronization delay to ensure the validity of the synchronization information of the primary and standby VNFs. Compared with other algorithms, the backup sharing method can effectively reduce operating expenses, improve the timeliness of the synchronization information between the primary and standby VNF instances, and has good robustness in ensuring the survivability of CSC. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic diagram of the service function chain in a traditional network;

[0019] Figure 2 Figure 1 compares two VNF backup methods. Figure a is deployment method 1 and figure b is deployment method 2.

[0020] Figure 3 A schematic diagram of a backup sharing example based on a capacity-constrained device location model of the present invention;

[0021] Figure 4 Schematic diagram of the joint game strategy of the present invention, where a is the master-slave mapping relationship output in stage 1, and b is the master-slave mapping relationship adjusted in stage 2;

[0022] Figure 5 This is a graph showing the results of the NSF network backup operation overhead experiment;

[0023] Figure 6 This is a graph showing the results of the USNet network backup operation overhead experiment;

[0024] Figure 7 This is the result of the NSF network synchronization information timeliness experiment;

[0025] Figure 8 This is the experimental result of USNet network synchronization information timeliness. DETAILED DESCRIPTION

[0026] In order to make the purpose, content and advantages of the present invention more clear, the specific implementation methods of the present invention are further described in detail below in conjunction with the drawings and examples.

[0027] Reserving redundant backup VNF resources can effectively guarantee the survivability of CSC in emergency situations. When a single node fails, the service traffic of the source CSC is adjusted to the backup VNF in a synchronized state for processing, thereby avoiding the impact of the time required for node failure recovery on service performance. However, it is worth noting that backup VNF resources also need to occupy the underlying computing, memory, and storage resources, thus introducing non-negligible redundant overhead. The present invention aims to design a reasonable backup VNF deployment method to minimize the backup redundancy and cut resource overhead while ensuring the survivability of CSC. Figure 2 As shown in the figure, for the same CSC request, backup method 1 backs up all VNFs, but it is obvious that the utilization rate of the underlying resources is too low, which is not conducive to operational benefits; since the backup resources in the cloud environment are shareable, backup method 2 adopts a shared backup method for the same VNF in the two CSCs. Under the premise that the synchronization delay constraint allows, this shared backup VNF deployment method can effectively compress the redundancy of backup resources and provide equivalent protection for single-node failures in the cloud environment.

[0028] The present invention first establishes a mathematical model for the important elements (including cloud environment elements and decision variables) that constitute the service chain in the cloud environment, and determines the constraints and optimization objectives of the cloud service chain CSC efficient backup (sharing) method:

[0029] 1.1 Cloud Environment Elements

[0030] (1) Underlying Network

[0031] The underlying network that provides computing resources can be modeled as an undirected weighted graph G = (V, E, A V ,A E ), where the underlying network node set is set to V and the attribute is A V ={C v ,L v}, where C v Indicates the maximum amount of computing resources that the underlying network node v can provide, L v Represents the relative position information of node v. The set of links connecting the underlying network nodes is set to E, and its attribute is A E ={λ e ,δ e}, where λ e represents the length of link e, δ e is the link synchronization communication delay.

[0032] (2) Virtual Network Function (VNF)

[0033] The active VNF instance m constitutes the set M, and the standby VNF instance Composition Collection The type of VNF is represented by the variable K, ck is the amount of computing resources required by k-type VNF instances, d k is the maximum number of times a k-type backup VNF instance is shared, which means that a k-type backup VNF instance can provide backup for at most several k-type active VNF instances. To ensure the synchronization of the active and standby states, the communication delay between the two shall not exceed the maximum synchronization delay of the k-type VNF δ k .

[0034] (3) Service Function Chain

[0035] Define I as a CSC set, a 3-element array A I ={s i ,o i ,β K(i)} is a set of CSC attributes. i ,o i ∈V are the source and destination nodes of service chain i respectively, β K(i) is the network function request set that constitutes the CSC, K(i) is the network function sequence of the CSC, such as

[0036] 1.2 Decision variables

[0037] Defining variables Indicates the primary and backup VNF instances Whether it is deployed on physical node v, Indicates whether the type of VNF instance m is k. Indicates a VNF instance Is the type of k, variable Indicates the active VNF instance m and the standby VNF instance Whether the state synchronization link between them passes through the link e(u,v), the decision variable Indicates the standby VNF instance Whether it is a backup of the active VNF instance m.

[0038] 1.3 Constraints

[0039] Both the active and standby VNF instances in the cloud environment consume computing resources. Therefore, the total computing resources required by the VNF instances deployed on any underlying network node must not exceed its maximum computing resources. Then:

[0040]

[0041] The same type of active VNF instances of the same underlying network node v constitute a set Due to the elastic expansion requirements of virtual network functions, Elements within There is usually a business dependency between them. To avoid business performance degradation caused by virtual resource separation, it is stipulated Cannot be split and backed up, that is, with the main VNF instance Corresponding standby instance They must also be concentrated on the same node:

[0042]

[0043] Among them, the same type of standby VNF instances of the same underlying network node v constitute a set To take full advantage of the backup resource sharing method, the The number of standby VNF instances of the same type in the table must not be less than any corresponding active VNF instance set. The number of elements in the constraint condition ensures that when any underlying network node fails, the underlying network node where the backup resource is located has the ability to carry the original resources:

[0044]

[0045] To ensure the survivability of the CSC in the event of a single node failure, the backup VNF instance and the active VNF instance cannot be located on the same node:

[0046]

[0047] Any node in the underlying network may be damaged, so each active VNF instance must have at least one standby VNF instance of the same type:

[0048]

[0049] The state synchronization capability of the standby VNF instance has a bottleneck. The number of times each standby VNF instance is backed up (shared) must not exceed the set maximum number of backups d. k :

[0050]

[0051] Finally, the states of the active and standby VNF instances must be synchronized to achieve effective backup. To ensure the timeliness of active and standby state synchronization, the communication delay between the two must be less than the set maximum synchronization delay δ k :

[0052]

[0053] Among them, δ e(u,v) represents the delay of link e(u,v).

[0054] 1.4 Optimization goals

[0055] The VNF efficient backup (sharing) method designed by the present invention aims to meet the constraints of CSC survivability. When a node failure occurs, the resources required for CSC service recovery are reduced by sharing the backup VNF instance, thereby reducing the cloud environment operation cost and reducing the communication delay between the primary and backup VNF instances as much as possible. The operation cost OPEX of the CSC survivability method includes the computing resources occupied by the backup VNF instance and the bandwidth resources occupied by the state synchronization of the primary and backup instances, where the computing resources are determined by the number of backup VNFs and the bandwidth resources depend on the logical length of the synchronization link. The OPEX can be expressed as formula (8), where r c , r b Represents the relative cost weight of computing resources and bandwidth resources:

[0056]

[0057] The synchronization state ST depends on the relative synchronization delay of the master and the backup k The proportion can be expressed as:

[0058]

[0059] In summary, the optimization objective of the VNF efficient backup (sharing) method involved in the present invention can be defined as formula (10), where the weight η∈(0,1):

[0060] min[(η)OPEX+(1-η)ST],st.Formula (1)-(7) (10)

[0061] In order to improve the benefits and performance of VNF backup resources, the present invention proposes an efficient backup method for service chains based on cloud environments. The method includes two stages, wherein stage 1 is an efficient backup deployment algorithm based on capacity-constrained device site selection, which abstracts the survivability problem of CSC into a site selection problem and compresses backup redundancy based on a taboo search algorithm; stage 2 uses a joint game strategy to optimize the relevance of the primary and backup virtual network function instances to ensure the validity of the synchronization state.

[0062] 2.1 Backup deployment algorithm based on capacity-constrained device location model (CFLP model)

[0063] The Capacitated Facility Location Problem (CFLP) aims to save transportation costs by optimizing equipment deployment. The deployment problem of backup VNF instances can be reduced to the CFLP problem. Figure 3 As shown in the figure, the underlying network nodes can be regarded as the candidate devices for providing services in the CFLP model, and the device service capacity is the maximum computing resource amount C of the node v, the equipment operation cost depends on the set of spare VNF instances The number of elements in; the set of active VNF instances on different nodes It can be regarded as users with different commodity processing needs, and the processing needs correspond to The number of elements in the CSC synchronization link logic length corresponds to the transportation overhead in the CFLP model. The survivable CSC efficient backup (sharing) (EBS) method needs to select several nodes from the underlying network nodes to carry the backup VNF instances, and establish a mapping relationship between the primary and backup VNF instances. This process is equivalent to using the CFLP model to select several devices from all devices to be open to users, and assign all user requirements to the selected devices to establish corresponding transportation paths. Note that the service capacity of the selected device in the CFLP model must not be less than the cumulative sum of user requirements. For the EBS method, when the backup VNF can be shared a sufficient number of times, if the set of standby VNF instances The number of elements is not less than the set of any active VNF instances associated with it If the number of elements is less than 1, there is no need to add a backup VNF instance. Figure 3 As shown, the number of backup VNF instances in the bottom node 5 and the bottom node 8 is not less than the number of active VNF instances in any node associated therewith, so backup (sharing) can be achieved. Therefore, it is necessary to improve on the basis of CFLP and design a solution algorithm suitable for the EBS method.

[0064] The evaluation conclusion of the reference literature shows that the taboo search algorithm has better results than other algorithms in solving the CFLP problem. Therefore, the present invention integrates the various constraints of the EBS method defined in 1.3 above and its commonality with the CFLP model, and designs an efficient backup deployment algorithm based on the capacity-constrained device site selection model to compress and backup VNF redundant resources. The process is shown in Algorithm 1, specifically as follows: First, input relevant information such as the underlying network nodes and links, and at the same time input the information that the main VNF set M has been deployed in the underlying network. Perform a full backup based on the distribution of the main VNF instances to initialize the minimum operating overhead OPEX min , the mapping matrix And taboo table The elements in are M and The operation cost under the mapping relationship is generated by traversing user needs and randomly assigning devices, under the premise of satisfying the device processing capacity and the device sharing constraints defined in 1.3 above. Then, the iterative optimization process is started. By continuously adjusting the mapping relationship between user requirements and devices in the original solution, the neighbor_set of the current feasible solution is generated, and the best feasible solution in the neighborhood is selected. If the optimal feasible solution does not meet the taboo search algorithm amnesty criteria, it is necessary to determine whether the operation that generated the optimal feasible solution is in the taboo table. When the operation is not in the taboo table (not subject to taboo constraints), retain the optimal feasible solution and update the operating expenditure OPEX min , when the operation is in the taboo table (subject to taboo constraints), the tabooed feasible solution is removed from the neighborhood neighbor_set, and the optimal feasible solution in the neighborhood is reselected for re-judgment. If the current optimal neighborhood optimal solution is better than the historical optimal feasible solution, then amnesty is enabled, that is, regardless of the optimal feasible solution generated Whether the operation is in the taboo table, the optimal feasible solution is accepted. Finally, the historical optimal feasible solution and the taboo table are updated and the next cycle is entered until the iteration ends and the OPEX is output. min and its corresponding

[0065]

[0066] 2.2 Joint game strategy

[0067] The efficient backup deployment algorithm based on the capacity-constrained device location model effectively reduces resource overhead by compressing the backup redundancy, but the timeliness of the preparation synchronization state still needs to be further optimized to avoid synchronization failure of the primary and backup VNF instances. Figure 4 As shown, Figure (a) represents the optimal output solution of the first stage This mapping relationship ignores the link delay in the process of optimizing the operation cost, which may cause excessive delay of some synchronization links. The modification is made to adjust the mapping relationship between the primary and backup instances in the public coverage domain of the backup VNF instances without changing the number of backup VNF instances, aiming to significantly improve the timeliness of the synchronization state by using a small amount of operating expenses OPEX. To describe the joint game strategy, the backup coverage domain and the backup coverage public domain are first defined.

[0068] Definition 1: Backup coverage area Define node n as belonging to the set of standby VNF instances Backup coverage area If and only if the shortest distance between nodes n and u does not exceed Maximum length of the synchronization link connected Right now in and

[0069] Definition 2: Backup covers public domain Define node n as belonging to a set of multiple standby VNF instances Backups of public domain If and only if the node n and the underlying network node set U = {u 1 ,u 2 …} the shortest distance between any nodes u does not exceed Maximum length of the synchronization link connected Right now in and

[0070] In order to balance the relationship between operational overhead and synchronization status timeliness, the joint game method adjusts the mapping relationship between the active VNF set covering the public domain and the original standby VNF instance set. Set by original backup VNF Remap to the new set of standby VNF instances There are two conditions that need to be met:

[0071] (1) Receiving The set of new standby VNF instances for the remap request Satisfy the maximum number of backup (sharing) times, that is

[0072] (2) New standby VNF set The total number of elements in the must not be less than the set of active VNF instances requesting remapping Right now

[0073] The joint game summary is shown in Algorithm 2, which is as follows: Each set of backup VNF instances calculates the backup coverage area (Steps 1-3), through Further calculations backup coverage public domain right The set of active VNF instances in Select a new set of standby VNF instances When the maximum number of backup (sharing) constraints and the backup resource carrying requirement constraints are met at the same time, the synchronization delay of the active and standby VNF instance sets before and after remapping is calculated. If the synchronization delay of the active and standby VNF instance sets after remapping is reduced, that is, Then update the mapping relationship between the primary and backup VNF instance sets (Step 9) This process is iterated until the set of all active VNF instances in the common domain of the backup coverage is traversed.

[0074]

[0075]

[0076] Specific embodiments of the present application are given below.

[0077] In order to evaluate the performance of the efficient backup sharing method, the present invention designs an experiment to compare with the local search algorithm (Local) and the BS-Pull algorithm. The evaluation indicators of the comparative experiment include operating overhead, synchronization information timeliness, and backup mechanism robustness. The simulation experiment platform is an Inter Core i7-4790 3.6GHz CPU, 8G memory, and a Linux system personal computer. The simulation environment is built using Python language, and the experimental results are analyzed with the help of Matlab.

[0078] The experiment selects two representative network topologies, NSF and USNet. The NSF network scale is small (30 nodes, 45 links) while the USNet is large (92 nodes, 152 links). Under the same network topology, the three algorithms are compared and analyzed by changing the scale of the active VNF instance set. The underlying network node carrying capacity follows U[200,300], the resource demand of each active VNF instance set follows U[10,50], the maximum number of times the standby VNF instance can be shared is set to 5, the link length follows U[50,750], and the link delay follows a random distribution of 100 to 300.

[0079] The simulation verification and analysis results are given below.

[0080] Experiment 1

[0081] This experiment shows that the operating costs of the three algorithms under different network scales are as follows: Figure 5 and Figure 6 As shown in the figure, it can be seen that as the number of active VNF instance sets increases, the operating overhead of various backup mechanisms shows an increasing trend. This is because the number of times that backup VNF instances can be shared is limited, and more backup VNF instances need to be deployed in the cloud environment to ensure the survivability of CSC. Compared with the Local algorithm and the BS-Pull algorithm, the efficient backup sharing method (EBS) can effectively reduce operating overhead through resource optimization. Figure 5 It can be seen that in the NSF network, when the number of active sets is small (40-70), the operating overhead based on the EBS mechanism is reduced by an average of 8.71% compared with the BS-Pull algorithm. When the number of active sets is 110-130, the operating overhead optimization effect can reach 11.18%. This is because the BS-Pull algorithm gives priority to nodes with the largest carrying domain, but as the number of active VNF sets increases, the difference in the carrying range of each node is small, and the optimization effect of selecting the largest carrying domain is no longer obvious. The Local algorithm is based on the idea of ​​​​hill climbing algorithm and is prone to fall into local optimal solutions. Therefore, the optimization effect of operating overhead is unstable. Figure 6It can be seen that the increase in network scale has little impact on the EBS algorithm and the BS-Pull algorithm. The EBS algorithm reduces the operating overhead by an average of 12.23% compared with the BS-Pull algorithm, while the gap between the Local algorithm and the two gradually increases. This is because the increase in network scale makes it easier for the Local algorithm to reside in the local optimal solution.

[0082] Experiment 2

[0083] This experiment compares the synchronization status effectiveness of different algorithms. According to the definition of synchronization delay effectiveness, the synchronization delay of the master and backup sets is relatively k The maximum value of the proportion directly reflects the timeliness of the synchronization information. Figure 7 It can be concluded that the active-standby synchronization delay ratio obtained by the EBS algorithm is less than 100%. As the number of active VNF instances increases, the active-standby synchronization delay ratio tends to decrease. This is because the more the backup instance set is required, the larger the scope of the backup coverage of the public domain, and the joint game strategy has more opportunities to adjust the active-standby mapping scheme. The BS-Pull algorithm only constrains the backup node bearer domain by the number of paths. Although this method plays a certain role in limiting the path length, it is unstable in terms of the effectiveness of synchronization information. Figure 8 It can be seen that as the problem scale increases, the proportion of the master-slave synchronization delay of the two algorithms will decrease. This is because as the number of deployed backup nodes increases, the backup node bearing domain and the backup coverage public domain will increase, and the synchronization link has more room for optimization.

[0084] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A service chain efficient backup method based on cloud environment, characterized in that: The following steps are involved: Step 1: Establish a mathematical model for the cloud environment elements and decision variables that constitute the service chain in the cloud environment, and determine the constraints and optimization goals of the cloud service chain CSC backup method; Step 2: Based on step 1, the backup deployment algorithm based on the capacity-constrained equipment site selection CFLP model is used to abstract the survivability problem of CSC into a site selection problem, and the backup redundancy is compressed based on the taboo search algorithm. Step 3: Based on step 2, a joint game strategy is used to optimize the association between the active and standby VNF instances to reduce the synchronization link delay between the active and standby VNF instances. In step 1, a mathematical model is established for cloud environment elements as follows: (1) Underlying Network The underlying network that provides computing resources is modeled as an undirected weighted graph G = (V, E, A V ,A E ), where the underlying network node set is set to V and the attribute is A V ={C v ,L v }, where C v Indicates the maximum amount of computing resources that the underlying network node v can provide, L v Represents the relative position information of node v; the set of links connecting the underlying network nodes is set to E, and its attribute is A E ={λ e ,δ e }, where λ e represents the length of link e, δ e is the link synchronization communication delay; (2) Virtual Network Function (VNF) The active VNF instance m constitutes the set M, and the standby VNF instance Composition Collection The type of VNF is represented by the variable K, c k is the amount of computing resources required by k-type VNF instances, d k is the maximum number of times a k-type backup VNF instance is shared, which means that a k-type backup VNF instance can provide backup for at most several k-type active VNF instances, and the communication delay between the active and backup VNF instances shall not exceed the maximum synchronization delay δ of the k-type VNF k ; (3) Service Function Chain Define I as a CSC set, a 3-element array A I ={s i ,o i ,β K(i) } is the CSC attribute set, s i ,o i ∈V are the source and destination nodes of service chain i respectively, β K(i) is the network function request set constituting the CSC, K(i) is the network function sequence of the CSC; In step 1, a mathematical model is established for the decision variables as follows: Define the variables Indicates the primary and backup VNF instances Whether it is deployed on physical node v, Indicates whether the type of VNF instance m is k. Indicates a VNF instance Is the type of k, variable Indicates the active VNF instance m and the standby VNF instance Whether the state synchronization link between them passes through the link e(u,v), the decision variable Indicates the standby VNF instance Whether it is a backup of the active VNF instance m; In step 1, the constraints of the cloud service chain CSC backup method are determined as follows: The total computing resources required by the VNF instances deployed on any underlying network node must not exceed its maximum computing resources, then: The same type of active VNF instances of the same underlying network node v constitute a set Elements within There is a business dependency relationship between Cannot be split and backed up, that is, with the main VNF instance Corresponding standby instance They must also be concentrated on the same node: lie in The number of standby VNF instances of the same type in the table must not be less than any corresponding active VNF instance set. The number of elements in : The backup VNF instance and the active VNF instance cannot be located on the same node: Each active VNF instance has at least one corresponding standby VNF instance of the same type: The number of times each standby VNF instance is backed up must not exceed the set maximum number of backups d k : The states of the active and standby VNF instances must be synchronized to achieve effective backup, and the communication delay between the two must be less than the set maximum synchronization delay δ k : Among them, δ e(u,v) represents the delay of link e(u,v); In step 1, the optimization objectives of the cloud service chain CSC backup method are determined as follows: The OPEX of the CSC survivability method includes the computing resources occupied by the backup VNF instance and the bandwidth resources occupied by the state synchronization of the primary and backup instances. The computing resources are determined by the number of backup VNFs, and the bandwidth resources depend on the logical length of the synchronization link. The OPEX is expressed as formula (8), where r c , r b Represents the relative cost weight of computing resources and bandwidth resources: The synchronization state ST depends on the relative synchronization delay of the master and the backup k The proportion of is expressed as: The optimization objective of the VNF efficient backup method can be defined as formula (10), where the weight η∈(0,1): min[(η)OPEX+(1-η)ST],st.(1)-(7) (10).

2. The method according to claim 1, characterized in that Step 2 is as follows: Based on the constraints of the CSC backup method defined in step 1 and its commonality with the CFLP model, a backup deployment algorithm based on the capacity-constrained device site selection model is designed to compress the backup VNF redundant resources. The process is as follows: first, input the information related to the underlying network nodes and links, and at the same time input the information that the active VNF set M has been deployed in the underlying network. Based on the distribution of the active VNF instances, a full backup is performed to initialize the minimum operating expenditure OPEX. min , the mapping matrix And taboo table The elements in are M and The operation cost under the mapping relationship is generated by traversing user needs and randomly assigning devices, under the premise of satisfying the device processing capacity and the constraints defined in step 1. Then, the iterative optimization process is started. By continuously adjusting the mapping relationship between user requirements and devices in the original solution, the neighbor_set of the current feasible solution is generated, and the best feasible solution in the neighborhood is selected. If the optimal feasible solution does not meet the taboo search algorithm amnesty criteria, it is determined whether the operation that generated the optimal feasible solution is in the taboo table. If the operation is not in the taboo table, the optimal feasible solution is retained and the operating expenditure OPEX is updated. min When the operation is in the taboo table, the taboo feasible solution is removed from the neighborhood neighbor_set, and the optimal feasible solution in the neighborhood is reselected for re-judgment. If the current optimal neighborhood optimal solution is better than the historical optimal feasible solution, the amnesty is enabled, that is, no matter what optimal feasible solution is generated The optimal feasible solution is accepted regardless of whether the operation is in the taboo table. Finally, the historical optimal feasible solution and the taboo table are updated and the next cycle is entered until the iteration ends and the OPEX is output. min and its corresponding 3. The method according to claim 2, characterized in that In step 3, the joint game strategy is used to Correction is made. Without changing the number of backup VNF instances, the mapping relationship between the primary and backup instances in the public coverage domain of the backup VNF instances is adjusted, and a certain amount of operating expenditure OPEX is used in exchange for a significant improvement in the timeliness of the synchronization status.

Images