WO2023184793A1 - Virtual network mapping algorithm based on delay optimization - Google Patents

Virtual network mapping algorithm based on delay optimization Download PDF

Info

Publication number
WO2023184793A1
WO2023184793A1 PCT/CN2022/106693 CN2022106693W WO2023184793A1 WO 2023184793 A1 WO2023184793 A1 WO 2023184793A1 CN 2022106693 W CN2022106693 W CN 2022106693W WO 2023184793 A1 WO2023184793 A1 WO 2023184793A1
Authority
WO
WIPO (PCT)
Prior art keywords
node
virtual
delay
mapping
physical
Prior art date
Application number
PCT/CN2022/106693
Other languages
French (fr)
Chinese (zh)
Inventor
陆音
杨超
孙君
亓晋
郭永安
Original Assignee
南京邮电大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 南京邮电大学 filed Critical 南京邮电大学
Priority to JP2023507645A priority Critical patent/JP2024515910A/en
Publication of WO2023184793A1 publication Critical patent/WO2023184793A1/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/02—Standardisation; Integration
    • H04L41/0226—Mapping or translating multiple network management protocols
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00—Packet switching elements
    • H04L49/70—Virtual switches

Definitions

  • the present invention relates to network virtualization technology and Software Defined Network (SDN, Software Defined Network) technology, specifically a virtual network mapping algorithm based on delay optimization.
  • SDN Software Defined Network
  • Network virtualization abstracts and encapsulates the existing physical resources of the network into virtual network elements that can be flexibly allocated and managed.
  • Network service providers can customize a variety of virtual networks as needed, and these virtual networks can exist independently on the underlying network without interfering with each other.
  • the physical network consists of physical nodes and physical links, and the physical resources involved include CPU, Node storage, node capacity, link bandwidth, etc.; the resource attribute requirements requested by the virtual network include CPU, node storage, node capacity, link bandwidth, etc., in addition to link delay requirements, node location requirements, physical distance Demand etc.
  • the current virtual network mapping algorithm consists of two sub-processes: virtual node mapping process and virtual link mapping process. Virtual network mapping is complete only after all virtual nodes and virtual links are mapped to underlying nodes. There are two main optimization goals for the virtual network mapping algorithm: 1. Minimize the cost of virtual network mapping and maximize the benefits of virtual network mapping; 2. Satisfy the network performance indicators QoS (Quality of Service) and QoE (Quality of Experience) of virtual network users ).
  • the present invention proposes a virtual virtualization method that adds link delay and node processing delay to the node sorting value in the node sorting algorithm for delay-sensitive application scenarios, and uses batch processing of virtual network requests within a time window.
  • Network mapping algorithm
  • the present invention is a virtual network mapping algorithm based on delay optimization, which includes the following steps:
  • Step 1 Establish a virtual network mapping model
  • Step 2 Sort the virtual nodes and physical nodes by importance based on the node's resource degree, the importance of the local topology, the propagation delay of the node's adjacent links, and the node processing delay;
  • Step 3 Perform virtual network node mapping
  • Step 4 Perform virtual link mapping.
  • step 2 the importance of virtual node n V is calculated as follows
  • Res(n V ), Closeness(n V ), and VertexDelay(n V ) are the resource concentration degree, node intimacy, and node propagation delay expectation of virtual node n V respectively;
  • n P The importance of physical node n P is calculated as follows
  • Res(n P ), Closeness(n P ), LocDis(n P ), and VertexDelay(n P ) respectively represent the resource concentration degree, node intimacy, local network topology importance, and node propagation time of the physical node n P. defer expectations;
  • step 3 is as follows:
  • Step 3.1 record the virtual nodes in the virtual network request into the VirtualNodeList in order of their importance from large to small;
  • Step 3.2 for the virtual node n V in the VirtualNodeList, traverse the underlying physical node set, select the candidate nodes that meet the physical distance constraints, and store them in the set Candidates(n V );
  • Step 3.3 determine whether the Candidates(n V ) set is empty. If it is empty, the mapping fails and the result is returned; if not, select the node with the highest node importance in the Candidates(n V ) set and map the virtual node n V Go to the candidate node, update the mapping relationship list MappingNodeList and the underlying physical resources, and delete the virtual node n V in the VirtualNodeList;
  • Step 3.4 repeat steps 3.2 and 3.3 until VirtualNodeList is empty.
  • step 4 the virtual link mapping adopts the K-Shortest path algorithm to select the shortest path and satisfy the bandwidth constraint.
  • the specific operations are as follows:
  • Step 4.1 sort the virtual links in the virtual network request from large to small and record them in VirtualLinkList;
  • Step 4.2 for the e V virtual link in VirtualLinkList, K candidate physical paths are found through the K-Shortest path algorithm, which is recorded as the set Paths (e V );
  • Step 4.3 determine the virtual link bandwidth requirement for the path in Paths(e V ). If the virtual link bandwidth requirement cannot be met, delete the path from Paths(e V );
  • Step 4.4 determine whether Paths (e V ) is empty. If it is empty, the mapping fails and the result is returned. If it is not empty, map the current virtual link e V to the physical link with the highest path priority and record the mapping relationship. Enter the collection MappingLinkList, delete the virtual link e V from the VirtualLinkList, and update the underlying physical resources;
  • Step 4.5 repeat steps 4.2 to 4.4 until VirtualLinkList is empty.
  • a further improvement of the present invention is that in step 4.4, the path priority is
  • bw(p) is the link bandwidth of the physical path p
  • ⁇ is the weight factor, which is taken as 1 in the present invention
  • hops(p) is the delay of the physical path p
  • Paths(e V ) is the virtual link e
  • the set of underlying physical paths after V mapping, p is a path in the set Paths(e V ).
  • the present invention adds the local topology value and resource concentration of the node to the node importance, and comprehensively considers the local and overall importance of the physical node in the underlying network, so that the virtual The physical node resources mapped by network nodes are more concentrated, thus improving its mapping success rate.
  • the present invention reflects the delay situation of adjacent links of nodes in the node importance, and can map virtual network nodes to low-latency physical nodes first; a delay constraint is added to the virtual link mapping stage, as shown in This improves the average network delay in QoS.
  • the simulation results show that this algorithm is better than the traditional algorithm in terms of QoS and has a higher mapping success rate.
  • Figure 1 is an example of virtual network mapping.
  • Figure 2 is a comparison chart of the mapping success rates of the three algorithms.
  • Figure 3 is a comparison chart of the long-term returns of the three algorithms.
  • Figure 4 is a comparison chart of the long-term benefit and expense ratios of the three algorithms.
  • Figure 5 is a comparison chart of the average propagation delay of the virtual network among the three algorithms.
  • the invention is a virtual network mapping algorithm based on delay optimization, which is divided into two stages; the node mapping part comprehensively considers the resource degree of the node, the importance of local topology, the propagation delay of the adjacent link of the node and the node processing time.
  • Yan proposed a new node importance ranking method; it can make the physical node resources mapped to virtual nodes more concentrated and the mapping success rate higher.
  • the link mapping stage the link propagation delay is used as the objective function, constraints are constructed based on the link propagation delay and link bandwidth, and the traditional dijkstra path algorithm is improved to map links.
  • the algorithm proposed by the present invention takes into account the success rate and long-term benefits of mapping, and improves virtual network QoS. Specific steps are as follows:
  • Step 1 Establish a virtual network mapping model
  • Step 2 Sort the importance of virtual nodes and physical nodes
  • Step 3 Perform virtual network node mapping
  • Step 4 Perform virtual link mapping.
  • the virtual network mapping model established in step 1 is as follows:
  • the attributes of physical node n P include CPU resource request cpu(n P ), node delay delay(n P ), and physical node importance(n p ).
  • the attributes of the physical link e P include link bandwidth bw(e P ) and link delay delay(e P ).
  • the resource request of virtual node n v is cpu(n V ), and the delay attribute is delay(n V ).
  • the bandwidth requirement of virtual link e v is bw(e V ), the delay requirement is delay(e V ), and the virtual node importance is importance(n V ).
  • a virtual network request consists of virtual nodes and virtual links.
  • the CPU resource requests of virtual network nodes a, b, c and each node are 10, 5, and 7 respectively; the link bandwidth requests between them are 15, 12, and 17.
  • the request can be mapped to nodes A, B, and C of the underlying physical network.
  • the essence of the virtual network mapping problem is to study how to effectively map virtual network requests to the underlying physical network.
  • Virtual network mapping is essentially an NP-hard problem, so finding an accurate optimal solution will take a lot of time and resources.
  • the mainstream solution to this problem is a heuristic algorithm, which is to find a feasible solution within a limited time.
  • This invention proposes a new heuristic algorithm.
  • the importance calculations of virtual node and physical node sorting are as follows:
  • ⁇ and ⁇ are weight factors used to adjust the proportion of CPU resources and bandwidth resources in the node resource degree.
  • ⁇ and ⁇ are both set to 1.
  • cpu(n) is the CPU resource of node n
  • bw(e) is the bandwidth resource of link e
  • Link(n) represents the set of links adjacent to node n.
  • Res(n) is determined by the sum of the CPU resources of node n and the available link bandwidth of its adjacent links, and can reflect the degree of resource concentration of node n in the entire network.
  • n i For any physical node or virtual node, record the node as n i , and the node intimacy reflects the topological importance of a node from the perspective of global topology.
  • n i and n j represent any two virtual nodes or two physical nodes. If n i is a virtual node, ⁇ (n i ) is a virtual node that has not been mapped in the virtual network; if n i is a physical node , ⁇ ( ni ) is a candidate node that satisfies conditional constraints in the physical network. hops(n i , n j ) is the hop distance between two nodes. The higher the node intimacy, the higher the centrality of the node in the network.
  • the parameter that measures the link delay around the node is the node propagation delay expectation, and its formula is as follows
  • delay(e i,j ) is the delay of virtual link e i,j
  • processDelay(n i ) is the processing delay of node n i
  • Degree(n i ) is the node degree of node n i , which represents The number of adjacent links for n i .
  • the numerator of equation (5) consists of the sum of link propagation delay and node processing delay, and describes the delay expectation of the network after the mapping is completed when passing through the node.
  • the propagation delay of each physical link and the processing delay of a single physical node are both 1 time unit, and the delay of a physical path is determined by the number of physical links included in the path.
  • n i in the node is a physical node, its expected node propagation delay parameter is
  • the selection priority of physical nodes is related to the physical distance.
  • candidate nodes that satisfy the conditional constraints candidate nodes that are close to the successfully mapped physical node should be given priority.
  • node n is any physical node
  • LocDis(n) is the local network topological importance of node n in the current network
  • N mapped is the physical node that has been occupied by virtual node mapping in the physical network
  • Dis(n, n j ) is the physical distance between nodes n and n j . The closer the virtual node is to the physical node after mapping, the fewer resources the mapping takes up and the higher the success rate of mapping.
  • the sub-process of virtual network node mapping is as follows:
  • Step 3.1 Sort the virtual node importance in the virtual network request calculated according to equation (1) from large to small and record it in VirtualNodeList;
  • Step 3.2 for the virtual node n V in the VirtualNodeList, traverse the underlying physical node set, select the candidate nodes that meet the physical distance constraints, and store them in the set Candidates(n V );
  • Step 3.3 determine whether the Candidates(n V ) set is empty. If it is empty, the mapping fails and the result is returned; if not, select the node importance calculated according to formula (2) in the Candidates(n V ) set. (n P ) The highest node, map the virtual node n V to the candidate node, update the mapping relationship list MappingNodeList and the underlying physical resources, and delete the virtual node n V in the VirtualNodeList;
  • Step 3.4 repeat steps 3.2 and 3.3 until VirtualNodeList is empty.
  • Virtual link mapping uses the K-Shortest path algorithm to select the shortest path and satisfy bandwidth constraints.
  • Step 4.1 sort the virtual links in the virtual network request from large to small and record them in VirtualLinkList;
  • Step 4.2 For the e V virtual link in the VirtualLinkList, K candidate physical paths are found through the K-Shortest path algorithm, which is recorded as the set Paths (e V ). In actual virtual network mapping, the value of K is affected by the scale of the underlying physical network. In the simulation below, the underlying physical network consists of 100 physical nodes and 500 physical links, and K is generally 5;
  • Step 4.3 determine the virtual link bandwidth requirement for the path in Paths(e V ). If the virtual link bandwidth requirement cannot be met, delete the path from Paths(e V );
  • Step 4.4 determine whether Paths (e V ) is empty. If it is empty, the mapping fails and the result is returned. If it is not empty, map the current virtual link e V to the physical link with the highest path priority and record the mapping relationship. Enter the collection MappingLinkList, delete the virtual link e V from the VirtualLinkList, and update the underlying physical resources;
  • Step 4.5 repeat steps 4.2 to 4.4 until VirtualLinkList is empty.
  • bw(p) is the bandwidth resource of path p
  • ⁇ is the weight factor, which is taken as 1 in the present invention
  • hops(p) is the hop distance of path p
  • Paths(e V ) is the virtual link e V mapping
  • the final set of candidate underlying physical paths, p is a physical path in the set Paths(e V ).
  • the hop count distance of the underlying physical path p represents the delay of this path.
  • the simulation parameters are set as follows:
  • the CPU resources of physical nodes and the bandwidth resources of physical links are uniformly distributed according to [50, 100]; the processing delay of physical nodes and the propagation delay of physical links are both 1 time unit; the propagation of virtual links
  • the delay is an integer and follows the uniform distribution of [1, 5].
  • Virtual network requests follow Poisson distribution, the arrival time is 100 time units, and the expected number is 5.
  • the survival time follows an exponential distribution with an expectation of 1000 time units.
  • the number of virtual network nodes follows a uniform distribution of [2, 10], the CPU resource requests of virtual network nodes and the bandwidth requests of virtual network links follow a uniform distribution of [0, 50], and the distance constraint of virtual nodes is 500.
  • the virtual network mapping indicators are as follows:
  • the mapping success rate can reflect the probability that a virtual network request is successfully mapped, as shown in Equation (9).
  • NUM suc is a successfully mapped virtual network request
  • NUM is a virtual network request that arrives within time T
  • ⁇ is a constant that approaches zero infinitely.
  • mapping revenue R(G V , t) and mapping cost C(G V , t) in time slot t are as follows
  • ⁇ and ⁇ are the weight factors of node resources and link bandwidth proportions, both of which are 1 in the present invention
  • cpu(n V ) is the CPU resource requirement of virtual node n V
  • bw(e V ) is the bandwidth resource requirement of virtual link e V
  • hops(p) is the number of hops of path p
  • p is a physical path in the set Paths(e V )
  • Paths(e V ) is the virtual link The set of underlying physical paths after e V mapping.
  • the long-term revenue-to-expense ratio can be expressed as
  • the numerator is the long-term mapping revenue of the virtual network request G V
  • the denominator is the long-term mapping cost of the virtual network request G V.
  • mapping revenue of a virtual network request is determined by the request itself and has nothing to do with the virtual network mapping algorithm used. Therefore, the performance of the virtual network mapping algorithm can be better seen from the long-term revenue-to-cost ratio. The closer the revenue-to-cost ratio is to 1, the The fewer underlying physical resources occupied by this mapping, the better the performance of the algorithm.
  • processDelay(n V ) is the delay of virtual node n V
  • delay(p) is the delay of path p corresponding to virtual link e V
  • NUM(N V ) and NUM(E V ) are the virtual network The number of virtual nodes and virtual links in GV .
  • the algorithm proposed in this invention is simulated and compared with the NC-VNE algorithm and CL algorithm.
  • the NC-VNE algorithm is a virtual network mapping algorithm that is reliably aware of nodes. This algorithm comprehensively considers the local topological importance of nodes, the resource degree of nodes, and the intimacy of nodes in the node sorting process.
  • the CL algorithm only considers the importance of node topology in node sorting. During simulation, delay constraints are added to the algorithm, and the comparison results are shown in Figures 2 to 5.
  • Figure 2 compares the virtual network request mapping success rates of the three algorithms. Since the algorithm proposed by the present invention comprehensively considers node resource degree, topological importance, node intimacy and adjacent link delay, its mapping success rate is the highest, and it tends to be stable after 7000 time units, and the mapping success rate is stable at 70 %, which is improved by about 33% compared to the NC-VNE algorithm. Since the CL algorithm does not consider parameters such as node resource degree, its mapping success rate is about 43%.
  • Figure 3 shows the comparison of the long-term average costs of the three algorithms.
  • the long-term average costs of the three algorithms all tend to stabilize after 5000 time units.
  • the long-term average return of the algorithm proposed in this invention is similar to the NC-VNE algorithm, and the long-term average return of the CL algorithm is the lowest. Compared with the CL algorithm, the long-term average income of the algorithm proposed by this invention is increased by about 32%.
  • Figure 4 shows the long-term average revenue-to-cost ratios of the three algorithms.
  • the long-term average revenue-to-cost ratios of the three algorithms tend to be stable after 5,000 units.
  • the long-term average revenue-to-cost ratio of the algorithm proposed by this invention is the highest, followed by the NC-VNE algorithm, and finally the CL algorithm.
  • Figure 5 shows the average network propagation delay of the three algorithms. As time goes by, the average network propagation delay of the three algorithms continues to increase and stabilizes after about 5,000 units. Since the algorithm proposed by the present invention considers link delay and node processing delay in node sorting and link mapping, nodes will be mapped to low-latency physical nodes first, and the results are better than the other two algorithms. Compared with the NC-VNE algorithm and the CL algorithm, the average propagation delay of the virtual network is reduced by 1 to 1.5 time units respectively.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

Disclosed in the present invention is a virtual network mapping algorithm based on delay optimization. The algorithm comprises the following steps: step 1, establishing a virtual network mapping model; step 2, performing importance sorting on a virtual node and a physical node; step 3, performing virtual network node mapping; and step 4, performing virtual link mapping. A node position constraint is taken into consideration, so as to reflect the importance of a node in a network topology. In the algorithm provided in the present invention, a link delay and a node processing delay are added to a node sorting value in a node sorting algorithm, and a method involving processing virtual network requests in batches within a time window is used; and compared with traditional algorithms, the algorithm is optimized in terms of an average network propagation delay and emphasizes the importance of a node in a local topology. A simulation result shows that the algorithm provided in the present invention is superior to traditional virtual network mapping algorithms in terms of QoS.

Description

基于时延优化的虚拟网络映射算法Virtual network mapping algorithm based on delay optimization 技术领域Technical field
本发明涉及网络虚拟化技术和软件定义网络(SDN，Software Defined Network)技术，具体地说是基于时延优化的虚拟网络映射算法。The present invention relates to network virtualization technology and Software Defined Network (SDN, Software Defined Network) technology, specifically a virtual network mapping algorithm based on delay optimization.
背景技术Background technique
面对当前互联网发展过程中遇到的僵化问题，学术界提出了以网络虚拟化技术为基础的网络切片解决方案。网络虚拟化将网络的现有物理资源抽象和封装为可灵活分配管理的虚拟网元。网络服务提供商可根据需要定制出各式各样的虚拟网络，这些虚拟网络可互不干扰地独立存在于底层网络。Faced with the rigidity problems encountered in the current development process of the Internet, academic circles have proposed network slicing solutions based on network virtualization technology. Network virtualization abstracts and encapsulates the existing physical resources of the network into virtual network elements that can be flexibly allocated and managed. Network service providers can customize a variety of virtual networks as needed, and these virtual networks can exist independently on the underlying network without interfering with each other.
在研究虚拟网络映射至底层物理网络的过程时，可以将虚拟网络和底层物理网络分别用无向有权图建模，其中物理网络由物理节点和物理链路构成，涉及的物理资源有CPU、节点存储、节点容量、链路带宽等；虚拟网络请求的资源属性需求包括CPU、节点存储、节点容量、链路带宽等，除此之外还有链路时延需求、节点位置需求、物理距离需求等。当前的虚拟网络映射算法由两个子过程组成：虚拟节点映射过程和虚拟链路映射过程。所有的虚拟节点和虚拟链路映射至底层节点后，虚拟网络映射才算完成。虚拟网络映射算法的优化目标主要有两个：1.最小化虚拟网络映射代价和最大化虚拟网络映射收益；2.满足虚拟网络用户的网络性能指标QoS(Quality of Service)和QoE(Quality of Experience)。When studying the process of mapping a virtual network to an underlying physical network, the virtual network and the underlying physical network can be modeled using undirected weighted graphs respectively. The physical network consists of physical nodes and physical links, and the physical resources involved include CPU, Node storage, node capacity, link bandwidth, etc.; the resource attribute requirements requested by the virtual network include CPU, node storage, node capacity, link bandwidth, etc., in addition to link delay requirements, node location requirements, physical distance Demand etc. The current virtual network mapping algorithm consists of two sub-processes: virtual node mapping process and virtual link mapping process. Virtual network mapping is complete only after all virtual nodes and virtual links are mapped to underlying nodes. There are two main optimization goals for the virtual network mapping algorithm: 1. Minimize the cost of virtual network mapping and maximize the benefits of virtual network mapping; 2. Satisfy the network performance indicators QoS (Quality of Service) and QoE (Quality of Experience) of virtual network users ).
如今，随着网络应用的多样化，各类应用对QoS的要求逐渐提高，其在各种低时延应用场景下，QoS中的虚拟网络平均传播时延需求更加严格。传统的虚拟网络映射算法主要通过节点排序、拓扑简化等方法实现最小化网络映射代价或最大化虚拟网络映射收益，忽略了对用户QoS的优化。此外，传统的两阶段协同虚拟网络映射算法主要考虑节点CPU资源和链路带宽，用以体现节点在整体网络中的资源集中度，而忽略了其他因素的影响。Nowadays, with the diversification of network applications, the requirements for QoS of various applications are gradually increasing. In various low-latency application scenarios, the requirements for the average propagation delay of the virtual network in QoS are more stringent. Traditional virtual network mapping algorithms mainly minimize network mapping costs or maximize virtual network mapping benefits through node sorting, topology simplification and other methods, ignoring the optimization of user QoS. In addition, the traditional two-stage collaborative virtual network mapping algorithm mainly considers node CPU resources and link bandwidth to reflect the node's resource concentration in the overall network, while ignoring the influence of other factors.
发明内容Contents of the invention
为了解决上述问题，本发明针对时延敏感型应用场景，提出了一种在节点排序算法中将链路时延和节点处理时延加入节点排序值，采用时间窗内批量处理虚拟网络请求的虚拟网络映射算法。In order to solve the above problems, the present invention proposes a virtual virtualization method that adds link delay and node processing delay to the node sorting value in the node sorting algorithm for delay-sensitive application scenarios, and uses batch processing of virtual network requests within a time window. Network mapping algorithm.
为了达到上述目的，本发明通过以下技术方案来实现：In order to achieve the above objects, the present invention is achieved through the following technical solutions:
本发明是基于时延优化的虚拟网络映射算法，包括以下步骤：The present invention is a virtual network mapping algorithm based on delay optimization, which includes the following steps:
步骤1，建立虚拟网络映射的模型；Step 1: Establish a virtual network mapping model;
步骤2，根据节点的资源度、局部拓扑的重要性、节点相邻链路的传播时延以及节点处理时延，对虚拟节点和物理节点进行重要度排序；Step 2: Sort the virtual nodes and physical nodes by importance based on the node's resource degree, the importance of the local topology, the propagation delay of the node's adjacent links, and the node processing delay;
步骤3，进行虚拟网络节点映射；Step 3: Perform virtual network node mapping;
步骤4，进行虚拟链路映射。Step 4: Perform virtual link mapping.
本发明的进一步改进在于：步骤1中虚拟网络映射的模型为：物理网络用无向有权图G P＝(N P，E P)表示，N P和E P分别为物理节点集合和物理链路集合，其中，物理节点n P的属性有CPU资源请求cpu(n P)、节点时延delay(n P)、物理节点重要度importance(n p)，物理链路e P的属性有链路带宽bw(e P)、链路时延delay(e P)； A further improvement of the present invention is that the virtual network mapping model in step 1 is: the physical network is represented by an undirected weighted graph G P = ( NP , EP ), where N P and EP are the physical node set and the physical chain respectively. path set, where the attributes of physical node n P include CPU resource request cpu(n P ), node delay delay(n P ), physical node importance(n p ), and the attributes of physical link e P include link Bandwidth bw(e P ), link delay delay(e P );
虚拟网络请求用无向有权图G V＝(N V，E V)表示，N V和E V分别为虚拟节点和虚拟链路集合，其中，虚拟节点n V的资源请求为cpu(n V)，时延属性为delay(n V)，虚拟链路e v的带宽约束为bw(e V)，时延约束为delay(e V)，虚拟节点重要度为importance(n V)。 The virtual network request is represented by an undirected weighted graph G V =(N V , EV ), where NV and EV are virtual nodes and virtual link sets respectively, where the resource request of virtual node n V is cpu(n V ), the delay attribute is delay(n V ), the bandwidth constraint of virtual link ev is bw(e V ), the delay constraint is delay(e V ), and the virtual node importance is importance(n V ).
本发明的进一步改进在于：步骤2中，虚拟节点n V的重要度计算如下 A further improvement of the present invention is that in step 2, the importance of virtual node n V is calculated as follows
Figure PCTCN2022106693-appb-000001
Figure PCTCN2022106693-appb-000001
式中，Res(n V)、Closeness(n V)、VertexDelay(n V)分别为虚拟节点n V的资源集中程度、节点亲密度、节点传播时延期望； In the formula, Res(n V ), Closeness(n V ), and VertexDelay(n V ) are the resource concentration degree, node intimacy, and node propagation delay expectation of virtual node n V respectively;
物理节点n P的重要度计算如下 The importance of physical node n P is calculated as follows
Figure PCTCN2022106693-appb-000002
Figure PCTCN2022106693-appb-000002
式中，Res(n P)、Closeness(n P)、LocDis(n P)、VertexDelay(n P)分别表示物理节点n P的资源集中程度、节点亲密度、局部网络拓扑重要性、节点传播时延期望； In the formula, Res(n P ), Closeness(n P ), LocDis(n P ), and VertexDelay(n P ) respectively represent the resource concentration degree, node intimacy, local network topology importance, and node propagation time of the physical node n P. defer expectations;
本发明的进一步改进在于：步骤3具体操作如下：A further improvement of the present invention is that the specific operation of step 3 is as follows:
步骤3.1，将虚拟网络请求中的虚拟节点按照其重要度从大到小的排序记录到VirtualNodeList中；Step 3.1, record the virtual nodes in the virtual network request into the VirtualNodeList in order of their importance from large to small;
步骤3.2，对于VirtualNodeList中的虚拟节点n V，遍历底层物理节点集合，选取满足物理距离约束条件的候选节点，存入集合Candidates(n V)； Step 3.2, for the virtual node n V in the VirtualNodeList, traverse the underlying physical node set, select the candidate nodes that meet the physical distance constraints, and store them in the set Candidates(n V );
步骤3.3，判断Candidates(n V)集合是否为空，若为空，映射失败，返回结果；若不为空，选择Candidates(n V)集合中节点重要度最高的节点，将虚拟节点n V映射至该候选节点上，更新映射关系列表MappingNodeList和底层物理资源，在VirtualNodeList中删除虚拟节点n V； Step 3.3, determine whether the Candidates(n V ) set is empty. If it is empty, the mapping fails and the result is returned; if not, select the node with the highest node importance in the Candidates(n V ) set and map the virtual node n V Go to the candidate node, update the mapping relationship list MappingNodeList and the underlying physical resources, and delete the virtual node n V in the VirtualNodeList;
步骤3.4，重复步骤3.2和步骤3.3，直至VirtualNodeList为空。Step 3.4, repeat steps 3.2 and 3.3 until VirtualNodeList is empty.
本发明的进一步改进在于：步骤4虚拟链路映射采用K-Shortest路径算法，选取最短路径并满足带宽约束，具体操作如下：A further improvement of the present invention is that in step 4, the virtual link mapping adopts the K-Shortest path algorithm to select the shortest path and satisfy the bandwidth constraint. The specific operations are as follows:
步骤4.1，将虚拟网络请求中的虚拟链路从大到小的排序记录到VirtualLinkList中；Step 4.1, sort the virtual links in the virtual network request from large to small and record them in VirtualLinkList;
步骤4.2，对于VirtualLinkList中的e V虚拟链路，通过K-Shortest路径算法寻找出K条候选物理路径，记作集合Paths(e V)； Step 4.2, for the e V virtual link in VirtualLinkList, K candidate physical paths are found through the K-Shortest path algorithm, which is recorded as the set Paths (e V );
步骤4.3，对Paths(e V)中的路径进行虚拟链路带宽需求判断，若不能满足虚拟链路带宽需求，则从Paths(e V)中删除该路径； Step 4.3, determine the virtual link bandwidth requirement for the path in Paths(e V ). If the virtual link bandwidth requirement cannot be met, delete the path from Paths(e V );
步骤4.4，判断Paths(e V)是否为空，若为空，映射失败，返回结果；若不为空，将当前虚拟链路e V映射至路径优先度最高的物理链路，将映射关系记录进集合MappingLinkList，从VirtualLinkList中删除虚拟链路e V，并更新底层物理资源； Step 4.4, determine whether Paths (e V ) is empty. If it is empty, the mapping fails and the result is returned. If it is not empty, map the current virtual link e V to the physical link with the highest path priority and record the mapping relationship. Enter the collection MappingLinkList, delete the virtual link e V from the VirtualLinkList, and update the underlying physical resources;
步骤4.5，重复步骤4.2至步骤4.4，直至VirtualLinkList为空。Step 4.5, repeat steps 4.2 to 4.4 until VirtualLinkList is empty.
本发明的进一步改进在于：步骤4.4中路径优先度为A further improvement of the present invention is that in step 4.4, the path priority is
Figure PCTCN2022106693-appb-000003
Figure PCTCN2022106693-appb-000003
式中，bw(p)为物理路径p的链路带宽，γ为权重因子，在本发明中取1，hops(p)为物理路径p的时延，Paths(e V)为虚拟链路e V映射后的底层物理路径的集合，p为集合Paths(e V)中的一条路径。 In the formula, bw(p) is the link bandwidth of the physical path p, γ is the weight factor, which is taken as 1 in the present invention, hops(p) is the delay of the physical path p, and Paths(e V ) is the virtual link e The set of underlying physical paths after V mapping, p is a path in the set Paths(e V ).
本发明的有益效果是：The beneficial effects of the present invention are:
1.本发明在虚拟网络映射算法中的节点映射子阶段中，在节点重要度中加入节点的局部拓扑值和资源集中度，综合考虑物理节点在底层网络中局部和整体的重要性，使得虚拟网络节点映射的物理节点资源更加集中，并因此提升其映射成功率。1. In the node mapping sub-stage of the virtual network mapping algorithm, the present invention adds the local topology value and resource concentration of the node to the node importance, and comprehensively considers the local and overall importance of the physical node in the underlying network, so that the virtual The physical node resources mapped by network nodes are more concentrated, thus improving its mapping success rate.
2.本发明在节点重要度中反映了节点相邻链路的时延情况，能够优先将虚拟 网络节点映射到低时延的物理节点；在虚拟链路映射阶段中加入了时延约束，由此改进了QoS中的网络平均时延。仿真结果表明，该算法在QoS方面优于传统算法，映射成功率更高。2. The present invention reflects the delay situation of adjacent links of nodes in the node importance, and can map virtual network nodes to low-latency physical nodes first; a delay constraint is added to the virtual link mapping stage, as shown in This improves the average network delay in QoS. The simulation results show that this algorithm is better than the traditional algorithm in terms of QoS and has a higher mapping success rate.
附图说明：Picture description:
图1是虚拟网络映射示例。Figure 1 is an example of virtual network mapping.
图2是3种算法的映射成功率比较图。Figure 2 is a comparison chart of the mapping success rates of the three algorithms.
图3是3种算法的长期收益比较图。Figure 3 is a comparison chart of the long-term returns of the three algorithms.
图4是3种算法的长期收益开销比比较图。Figure 4 is a comparison chart of the long-term benefit and expense ratios of the three algorithms.
图5是3种算法的虚拟网络平均传播时延比较图。Figure 5 is a comparison chart of the average propagation delay of the virtual network among the three algorithms.
具体实施方式Detailed ways
下面将结合本发明中的附图，对本发明实施例中的技术方案进行清楚、完整的描述。显然，所描述的实施例仅仅是本发明的一部分实施，而不是全部的实施。基于本发明的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得所有其他实施例，都属于本发明保护的范围。The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are only part of the implementations of the present invention, but not all implementations. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.
本发明是一种基于时延优化的虚拟网络映射算法，分为两阶段；节点映射部分综合考虑了节点的资源度、局部拓扑的重要性、节点相邻链路的传播时延以及节点处理时延，提出了一种新的节点重要度排序方法；可以使得虚拟节点映射至的物理节点资源更加集中，映射成功率更高。在链路映射阶段，以链路传播时延为目标函数，基于链路传播时延和链路带宽构建约束，改进传统的dijkstra路径算法映射链路。经过仿真验证，本发明所提算法兼顾了映射的成功率和长期收益，改进了虚拟网络QoS。具体步骤如下：The invention is a virtual network mapping algorithm based on delay optimization, which is divided into two stages; the node mapping part comprehensively considers the resource degree of the node, the importance of local topology, the propagation delay of the adjacent link of the node and the node processing time. Yan, proposed a new node importance ranking method; it can make the physical node resources mapped to virtual nodes more concentrated and the mapping success rate higher. In the link mapping stage, the link propagation delay is used as the objective function, constraints are constructed based on the link propagation delay and link bandwidth, and the traditional dijkstra path algorithm is improved to map links. After simulation verification, the algorithm proposed by the present invention takes into account the success rate and long-term benefits of mapping, and improves virtual network QoS. Specific steps are as follows:
步骤1，建立虚拟网络映射的模型；Step 1: Establish a virtual network mapping model;
步骤2，对虚拟节点和物理节点进行重要度排序；Step 2: Sort the importance of virtual nodes and physical nodes;
步骤3，进行虚拟网络节点映射；Step 3: Perform virtual network node mapping;
步骤4，进行虚拟链路映射。Step 4: Perform virtual link mapping.
步骤1中建立的虚拟网络映射模型如下：The virtual network mapping model established in step 1 is as follows:
物理网络用无向有权图G P＝(N P，E P)表示，N P和E P分别为物理节点集合和物理链路集合。物理节点n P的属性有CPU资源请求cpu(n P)、节点时延delay(n P)、物理节点重要度importance(n p)。物理链路e P的属性有链路带宽bw(e P)、链路 时延delay(e P)。虚拟网络请求用无向有权图G V＝(N V，E V)表示，N V和E V分别为虚拟节点和虚拟链路集合。虚拟节点n v的资源请求为cpu(n V)，时延属性为delay(n V)。虚拟链路e v的带宽需求为bw(e V)，时延需求为delay(e V)，虚拟节点重要度importance(n V)。 The physical network is represented by an undirected weighted graph G P =( NP , EP ), where N P and EP are the physical node set and the physical link set respectively. The attributes of physical node n P include CPU resource request cpu(n P ), node delay delay(n P ), and physical node importance(n p ). The attributes of the physical link e P include link bandwidth bw(e P ) and link delay delay(e P ). The virtual network request is represented by an undirected weighted graph G V =( NV , EV ), where NV and EV are virtual nodes and virtual link sets respectively. The resource request of virtual node n v is cpu(n V ), and the delay attribute is delay(n V ). The bandwidth requirement of virtual link e v is bw(e V ), the delay requirement is delay(e V ), and the virtual node importance is importance(n V ).
如图1所示，一个虚拟网络请求由虚拟节点和虚拟链路组成。其虚拟网络节点a、b、c和各个节点的CPU资源请求分别为10、5、7；它们之间的链路带宽请求为15、12、17。在满足了CPU资源约束和带宽约束的情况下，该请求可以映射至底层物理网络的A、B、C节点。虚拟网络映射问题本质就是研究如何有效地将虚拟网络请求映射至底层物理网络之上。As shown in Figure 1, a virtual network request consists of virtual nodes and virtual links. The CPU resource requests of virtual network nodes a, b, c and each node are 10, 5, and 7 respectively; the link bandwidth requests between them are 15, 12, and 17. When the CPU resource constraints and bandwidth constraints are met, the request can be mapped to nodes A, B, and C of the underlying physical network. The essence of the virtual network mapping problem is to study how to effectively map virtual network requests to the underlying physical network.
虚拟网络映射本质是一个NP-hard问题，因此寻找出精确的最优解将花费大量的时间和资源。目前，该问题的主流解决方法为启发式算法，即在有限时间内寻找出一个可行解。Virtual network mapping is essentially an NP-hard problem, so finding an accurate optimal solution will take a lot of time and resources. Currently, the mainstream solution to this problem is a heuristic algorithm, which is to find a feasible solution within a limited time.
本发明提出一种新型启发式算法，虚拟节点和物理节点排序的重要度计算分别如下This invention proposes a new heuristic algorithm. The importance calculations of virtual node and physical node sorting are as follows:
Figure PCTCN2022106693-appb-000004
Figure PCTCN2022106693-appb-000004
Figure PCTCN2022106693-appb-000005
Figure PCTCN2022106693-appb-000005
在式(1)和式(2)中，对于任意一个物理节点或虚拟节点n，反映该节点的资源集中程度的参数如下In equations (1) and (2), for any physical node or virtual node n, the parameters reflecting the resource concentration degree of the node are as follows:
Figure PCTCN2022106693-appb-000006
Figure PCTCN2022106693-appb-000006
式中，α和β是权重因子，用以调整CPU资源和带宽资源在节点资源度中占的比重，在本发明中α和β均取1。cpu(n)为节点n的CPU资源，bw(e)为链路e的带宽资源，Link(n)表示节点n邻接的链路集合。Res(n)由节点n的CPU资源和与其相邻链路的可用链路带宽之和决定，能够体现节点n在整个网络中的资源集中程度。In the formula, α and β are weight factors used to adjust the proportion of CPU resources and bandwidth resources in the node resource degree. In the present invention, α and β are both set to 1. cpu(n) is the CPU resource of node n, bw(e) is the bandwidth resource of link e, and Link(n) represents the set of links adjacent to node n. Res(n) is determined by the sum of the CPU resources of node n and the available link bandwidth of its adjacent links, and can reflect the degree of resource concentration of node n in the entire network.
对于任意一个物理节点或虚拟节点，将该节点记为n i，节点亲密度从全局拓扑的角度反映一个节点拓扑上的重要性。 For any physical node or virtual node, record the node as n i , and the node intimacy reflects the topological importance of a node from the perspective of global topology.
Figure PCTCN2022106693-appb-000007
Figure PCTCN2022106693-appb-000007
式中，n i与n j代表任意两个虚拟节点或两个物理节点，若n i为虚拟节点，Φ(n i)为在虚拟网络中还未映射的虚拟节点；若n i为物理节点，Φ(n i)为在物理网络中满足条件约束的候选节点。hops(n i，n j)为两个节点间的跳数距离，节点亲密度越高，该节点在该网络中的中心程度越高。 In the formula, n i and n j represent any two virtual nodes or two physical nodes. If n i is a virtual node, Φ(n i ) is a virtual node that has not been mapped in the virtual network; if n i is a physical node , Φ( ni ) is a candidate node that satisfies conditional constraints in the physical network. hops(n i , n j ) is the hop distance between two nodes. The higher the node intimacy, the higher the centrality of the node in the network.
对于任意一个物理节点或虚拟节点n，衡量该节点周边链路时延的参数为节点传播时延期望，其公式如下For any physical node or virtual node n, the parameter that measures the link delay around the node is the node propagation delay expectation, and its formula is as follows
Figure PCTCN2022106693-appb-000008
Figure PCTCN2022106693-appb-000008
式中，delay(e i，j)为虚拟链路e i，j的时延，processDelay(n i)为节点n i的处理时延，Degree(n i)为节点n i的节点度，代表n i的邻接链路数量。式(5)的分子由链路传播时延和节点处理时延之和组成，描述的是映射完成后的网络在经过该节点时的时延期望。 In the formula, delay(e i,j ) is the delay of virtual link e i,j , processDelay(n i ) is the processing delay of node n i , and Degree(n i ) is the node degree of node n i , which represents The number of adjacent links for n i . The numerator of equation (5) consists of the sum of link propagation delay and node processing delay, and describes the delay expectation of the network after the mapping is completed when passing through the node.
在本发明中，每条物理链路的传播时延和单个物理节点的处理时延均为1个时间单位，物理路径的时延由该路径中所包含的物理链路数量决定。当节点中的n i为物理节点时，其节点传播时延期望参数为 In the present invention, the propagation delay of each physical link and the processing delay of a single physical node are both 1 time unit, and the delay of a physical path is determined by the number of physical links included in the path. When n i in the node is a physical node, its expected node propagation delay parameter is
Figure PCTCN2022106693-appb-000009
Figure PCTCN2022106693-appb-000009
在虚拟网络节点映射过程中，物理节点的选取优先级与物理距离有关。在满足条件约束的候选节点中，应优先选取距离已成功映射的物理节点近的候选节点。In the virtual network node mapping process, the selection priority of physical nodes is related to the physical distance. Among the candidate nodes that satisfy the conditional constraints, candidate nodes that are close to the successfully mapped physical node should be given priority.
Figure PCTCN2022106693-appb-000010
Figure PCTCN2022106693-appb-000010
式中，节点n为任意物理节点，LocDis(n)为节点n在当前网络的局部网络拓扑重要性，N mapped为在物理网络中已被虚拟节点映射占用的物理节点，Dis(n，n j)为节点n和n j之间的物理距离。虚拟节点映射后的物理节点距离越近，映射所占用的资源越少，映射的成功率也更高。 In the formula, node n is any physical node, LocDis(n) is the local network topological importance of node n in the current network, N mapped is the physical node that has been occupied by virtual node mapping in the physical network, Dis(n, n j ) is the physical distance between nodes n and n j . The closer the virtual node is to the physical node after mapping, the fewer resources the mapping takes up and the higher the success rate of mapping.
虚拟网络节点映射子过程如下：The sub-process of virtual network node mapping is as follows:
步骤3.1，将根据式(1)计算得到的虚拟网络请求中的虚拟节点重要度从大到小的排序记录到VirtualNodeList中；Step 3.1: Sort the virtual node importance in the virtual network request calculated according to equation (1) from large to small and record it in VirtualNodeList;
步骤3.2，对于VirtualNodeList中的虚拟节点n V，遍历底层物理节点集合，选取满足物理距离约束条件的候选节点，存入集合Candidates(n V)； Step 3.2, for the virtual node n V in the VirtualNodeList, traverse the underlying physical node set, select the candidate nodes that meet the physical distance constraints, and store them in the set Candidates(n V );
步骤3.3，判断Candidates(n V)集合是否为空，若为空，映射失败，返回结果；若不为空，选择Candidates(n V)集合中的根据式(2)计算得到的节点重要度importance(n P)最高的节点，将虚拟节点n V映射至该候选节点上，并更新映射关系列表MappingNodeList和底层物理资源，在VirtualNodeList中删除虚拟节点n V； Step 3.3, determine whether the Candidates(n V ) set is empty. If it is empty, the mapping fails and the result is returned; if not, select the node importance calculated according to formula (2) in the Candidates(n V ) set. (n P ) The highest node, map the virtual node n V to the candidate node, update the mapping relationship list MappingNodeList and the underlying physical resources, and delete the virtual node n V in the VirtualNodeList;
步骤3.4，重复步骤3.2和步骤3.3，直至VirtualNodeList为空。Step 3.4, repeat steps 3.2 and 3.3 until VirtualNodeList is empty.
成功映射虚拟网络所有节点后，将虚拟链路映射至底层物理网络上。虚拟链路映射采用K-Shortest路径算法，选取最短路径并满足带宽约束。After successfully mapping all nodes of the virtual network, map the virtual links to the underlying physical network. Virtual link mapping uses the K-Shortest path algorithm to select the shortest path and satisfy bandwidth constraints.
虚拟链路映射的过程如下：The process of virtual link mapping is as follows:
步骤4.1，将虚拟网络请求中的虚拟链路从大到小的排序记录到VirtualLinkList中；Step 4.1, sort the virtual links in the virtual network request from large to small and record them in VirtualLinkList;
步骤4.2，对于VirtualLinkList中的e V虚拟链路，通过K-Shortest路径算法寻找出K条候选物理路径，记作集合Paths(e V)。在实际虚拟网络映射中，K的值受底层物理网络的规模影响。在下文的仿真中，底层物理网络由100个物理节点与500条物理链路组成，K一般取5； Step 4.2: For the e V virtual link in the VirtualLinkList, K candidate physical paths are found through the K-Shortest path algorithm, which is recorded as the set Paths (e V ). In actual virtual network mapping, the value of K is affected by the scale of the underlying physical network. In the simulation below, the underlying physical network consists of 100 physical nodes and 500 physical links, and K is generally 5;
步骤4.3，对Paths(e V)中的路径进行虚拟链路带宽需求判断，若不能满足虚拟链路带宽需求，则从Paths(e V)中删除该路径； Step 4.3, determine the virtual link bandwidth requirement for the path in Paths(e V ). If the virtual link bandwidth requirement cannot be met, delete the path from Paths(e V );
步骤4.4，判断Paths(e V)是否为空，若为空，映射失败，返回结果；若不为空，将当前虚拟链路e V映射至路径优先度最高的物理链路，将映射关系记录进集合MappingLinkList，从VirtualLinkList中删除虚拟链路e V，并更新底层物理资源； Step 4.4, determine whether Paths (e V ) is empty. If it is empty, the mapping fails and the result is returned. If it is not empty, map the current virtual link e V to the physical link with the highest path priority and record the mapping relationship. Enter the collection MappingLinkList, delete the virtual link e V from the VirtualLinkList, and update the underlying physical resources;
步骤4.5，重复步骤4.2至步骤4.4，直至VirtualLinkList为空。Step 4.5, repeat steps 4.2 to 4.4 until VirtualLinkList is empty.
对于集合Paths(e V)中的任意一条路径p，综合考虑链路带宽和链路时延，将其总结为参数路径优先度，并由此决定优先映射的路径，计算公式如下 For any path p in the set Paths(e V ), considering the link bandwidth and link delay, it is summarized as the parameter path priority, and the path to be mapped first is determined accordingly. The calculation formula is as follows
Figure PCTCN2022106693-appb-000011
Figure PCTCN2022106693-appb-000011
式中，bw(p)为路径p的带宽资源，γ为权重因子，在本发明中取1，hops(p)为路径p的跳数距离，Paths(e V)为虚拟链路e V映射后的候选底层物理路径集合，p为 集合Paths(e V)中的一条物理路径。底层物理路径p的跳数距离代表了该条路径的时延。 In the formula, bw(p) is the bandwidth resource of path p, γ is the weight factor, which is taken as 1 in the present invention, hops(p) is the hop distance of path p, and Paths(e V ) is the virtual link e V mapping The final set of candidate underlying physical paths, p is a physical path in the set Paths(e V ). The hop count distance of the underlying physical path p represents the delay of this path.
仿真参数设置如下：The simulation parameters are set as follows:
将100个物理节点和500个物理链路随机分布在1000×1000的范围内。其中，物理节点的CPU资源和物理链路的带宽资源均遵从[50，100]均匀分布；物理节点的处理时延和物理链路的传播时延均为1个时间单位；虚拟链路的传播时延为整数，遵从[1，5]的均匀分布。虚拟网络请求遵从泊松分布，到达时间为100个时间单位，期望个数为5。生存时间遵从指数分布，期望为1000个时间单位。虚拟网络节点个数遵从[2，10]均匀分布，虚拟网络节点的CPU资源请求和虚拟网络链路的带宽请求均遵从[0，50]均匀分布，虚拟节点的距离约束为500。Randomly distribute 100 physical nodes and 500 physical links within a range of 1000×1000. Among them, the CPU resources of physical nodes and the bandwidth resources of physical links are uniformly distributed according to [50, 100]; the processing delay of physical nodes and the propagation delay of physical links are both 1 time unit; the propagation of virtual links The delay is an integer and follows the uniform distribution of [1, 5]. Virtual network requests follow Poisson distribution, the arrival time is 100 time units, and the expected number is 5. The survival time follows an exponential distribution with an expectation of 1000 time units. The number of virtual network nodes follows a uniform distribution of [2, 10], the CPU resource requests of virtual network nodes and the bandwidth requests of virtual network links follow a uniform distribution of [0, 50], and the distance constraint of virtual nodes is 500.
仿真运行10000个时间单位，为去除仿真的随机因素，取10次运行的平均值。The simulation runs for 10,000 time units. In order to remove the random factors of the simulation, the average of 10 runs is taken.
虚拟网络映射指标如下：The virtual network mapping indicators are as follows:
映射成功率能反映出虚拟网络请求成功被映射的概率，如式(9)所示。The mapping success rate can reflect the probability that a virtual network request is successfully mapped, as shown in Equation (9).
Figure PCTCN2022106693-appb-000012
Figure PCTCN2022106693-appb-000012
式中，NUM suc为成功映射的虚拟网络请求，NUM为在时间T内到达的虚拟网络请求，δ为无限趋近于零的常数。 In the formula, NUM suc is a successfully mapped virtual network request, NUM is a virtual network request that arrives within time T, and δ is a constant that approaches zero infinitely.
对于虚拟网络G V＝(N V，E V)，在t时隙的映射收益R(G V，t)和映射开销C(G V，t)如下 For the virtual network G V =(N V , E V ), the mapping revenue R(G V , t) and mapping cost C(G V , t) in time slot t are as follows
Figure PCTCN2022106693-appb-000013
Figure PCTCN2022106693-appb-000013
Figure PCTCN2022106693-appb-000014
Figure PCTCN2022106693-appb-000014
在式(10)和式(11)中，λ和ω为节点资源和链路带宽比重的权重因子，在本发明中均为1；cpu(n V)为虚拟节点n V的CPU资源需求，bw(e V)为虚拟链路e V的带宽资源需求，hops(p)为路径p的跳数，p为集合Paths(e V)中的一条物理路径，Paths(e V)为虚拟链路e V映射后的底层物理路径集合。 In equations (10) and (11), λ and ω are the weight factors of node resources and link bandwidth proportions, both of which are 1 in the present invention; cpu(n V ) is the CPU resource requirement of virtual node n V , bw(e V ) is the bandwidth resource requirement of virtual link e V , hops(p) is the number of hops of path p, p is a physical path in the set Paths(e V ), and Paths(e V ) is the virtual link The set of underlying physical paths after e V mapping.
长期收益开销比可表示为The long-term revenue-to-expense ratio can be expressed as
Figure PCTCN2022106693-appb-000015
Figure PCTCN2022106693-appb-000015
式中，分子为虚拟网络请求G V的长期映射收益，分母为虚拟网络请求G V的长期映射开销。 In the formula, the numerator is the long-term mapping revenue of the virtual network request G V , and the denominator is the long-term mapping cost of the virtual network request G V.
一个虚拟网络请求的映射收益是由请求本身决定的，与使用的虚拟网络映射算法无关，因此从长期收益开销比能更好的看出虚拟网络映射算法的性能，收益开销比越接近1，则此次映射占用的底层物理资源越少，该算法的性能越好。The mapping revenue of a virtual network request is determined by the request itself and has nothing to do with the virtual network mapping algorithm used. Therefore, the performance of the virtual network mapping algorithm can be better seen from the long-term revenue-to-cost ratio. The closer the revenue-to-cost ratio is to 1, the The fewer underlying physical resources occupied by this mapping, the better the performance of the algorithm.
对于一个虚拟网络请求G V＝(N V，E V)，虚拟网络平均传播时延AveDelay(G V)为 For a virtual network request G V =(N V , EV ), the virtual network average propagation delay AveDelay(G V ) is
Figure PCTCN2022106693-appb-000016
Figure PCTCN2022106693-appb-000016
式中，processDelay(n V)为虚拟节点n V的时延，delay(p)为虚拟链路e V所对应的路径p的时延，NUM(N V)和NUM(E V)为虚拟网络G V中虚拟节点和虚拟链路的数量。 In the formula, processDelay(n V ) is the delay of virtual node n V , delay(p) is the delay of path p corresponding to virtual link e V , NUM(N V ) and NUM(E V ) are the virtual network The number of virtual nodes and virtual links in GV .
仿真结果如下：The simulation results are as follows:
将本发明所提算法与NC-VNE算法和CL算法进行仿真对比。NC-VNE算法为节点可靠感知的虚拟网络映射算法，该算法在节点排序过程中综合考虑了节点的局部拓扑重要性、节点的资源度和节点的亲密度。CL算法在节点排序中只考虑了节点拓扑的重要性。仿真时，在算法中加入时延约束，其对比结果如图2至图5所示。The algorithm proposed in this invention is simulated and compared with the NC-VNE algorithm and CL algorithm. The NC-VNE algorithm is a virtual network mapping algorithm that is reliably aware of nodes. This algorithm comprehensively considers the local topological importance of nodes, the resource degree of nodes, and the intimacy of nodes in the node sorting process. The CL algorithm only considers the importance of node topology in node sorting. During simulation, delay constraints are added to the algorithm, and the comparison results are shown in Figures 2 to 5.
图2比较了3种算法的虚拟网络请求映射成功率。本发明所提算法由于综合考虑了节点资源度、拓扑重要性、节点亲密度以及相邻链路时延，其映射成功率最高，在7000个时间单位后趋于稳定，映射成功率稳定在70％，相较NC-VNE算法提高了约33％。CL算法由于未考虑节点资源度等参数，其映射成功率在43％左右。Figure 2 compares the virtual network request mapping success rates of the three algorithms. Since the algorithm proposed by the present invention comprehensively considers node resource degree, topological importance, node intimacy and adjacent link delay, its mapping success rate is the highest, and it tends to be stable after 7000 time units, and the mapping success rate is stable at 70 %, which is improved by about 33% compared to the NC-VNE algorithm. Since the CL algorithm does not consider parameters such as node resource degree, its mapping success rate is about 43%.
图3中为3种算法的长期平均开销对比，3种算法的长期平均开销均在5000个时间单位后趋于稳定。本发明所提算法的长期平均收益与NC-VNE算法相近，CL算法的长期平均收益最低。本发明所提算法相较CL算法，其长期平均收益提升了32％左右。Figure 3 shows the comparison of the long-term average costs of the three algorithms. The long-term average costs of the three algorithms all tend to stabilize after 5000 time units. The long-term average return of the algorithm proposed in this invention is similar to the NC-VNE algorithm, and the long-term average return of the CL algorithm is the lowest. Compared with the CL algorithm, the long-term average income of the algorithm proposed by this invention is increased by about 32%.
图4为3种算法的长期平均收益开销比，3种算法的长期平均收益开销比在5000个单位后趋于稳定。本发明所提算法的长期平均收益开销比最高，其次为NC-VNE算法，最后为CL算法。Figure 4 shows the long-term average revenue-to-cost ratios of the three algorithms. The long-term average revenue-to-cost ratios of the three algorithms tend to be stable after 5,000 units. The long-term average revenue-to-cost ratio of the algorithm proposed by this invention is the highest, followed by the NC-VNE algorithm, and finally the CL algorithm.
图5为3种算法的网络平均传播时延。随着时间的推移，3种算法的网络平均传播时延不断延长，并在约5000个单位后趋于稳定。由于本发明所提算法在节点排序和链路映射中考虑了链路时延与节点处理时延，因此会优先将节点映射至低时延的物理节点，其结果优于其他两种算法。虚拟网络平均传播时延相较NC-VNE算法和CL算法分别减少了1至1.5个时间单位。Figure 5 shows the average network propagation delay of the three algorithms. As time goes by, the average network propagation delay of the three algorithms continues to increase and stabilizes after about 5,000 units. Since the algorithm proposed by the present invention considers link delay and node processing delay in node sorting and link mapping, nodes will be mapped to low-latency physical nodes first, and the results are better than the other two algorithms. Compared with the NC-VNE algorithm and the CL algorithm, the average propagation delay of the virtual network is reduced by 1 to 1.5 time units respectively.
以上所述仅是本发明的优选实施例而已，并非对本发明做任何形式上的限制。虽然本发明已以优选实施例论述如上，然而并非用以限定本发明，任何熟悉本专业的技术人员，在不脱离本发明技术方案的范围内，当可利用上述揭示的技术内容做出些许更动或修饰为等同变化的等效实施例。但凡是未脱离本发明技术方案的内容，依据本实用发明的技术实质对以上实施例所作的任何简单修改、等同变化与修饰，均仍属于本发明技术方案的范围内。The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention in any form. Although the present invention has been discussed above in terms of preferred embodiments, this is not intended to limit the present invention. Any skilled person familiar with the art can make some modifications using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. Changes or modifications are equivalent embodiments of equivalent changes. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims (6)

  1. 基于时延优化的虚拟网络映射算法，其特征在于：包括以下步骤：The virtual network mapping algorithm based on delay optimization is characterized by: including the following steps:
    步骤1，建立虚拟网络映射的模型；Step 1: Establish a virtual network mapping model;
    步骤2，根据节点的资源度、局部拓扑的重要性、节点相邻链路的传播时延以及节点处理时延，对虚拟节点和物理节点进行重要度排序；Step 2: Sort the virtual nodes and physical nodes by importance based on the node's resource degree, the importance of the local topology, the propagation delay of the node's adjacent links, and the node processing delay;
    步骤3，进行虚拟网络节点映射；Step 3: Perform virtual network node mapping;
    步骤4，进行虚拟链路映射。Step 4: Perform virtual link mapping.
  2. 根据权利要求1所述基于时延优化的虚拟网络映射算法，其特征在于：步骤1中虚拟网络映射的模型为：物理网络用无向有权图G P＝(N P,E P)表示，N P和E P分别为物理节点集合和物理链路集合，其中，物理节点n P的属性有CPU资源请求cpu(n P)、节点时延delay(n P)、物理节点重要度importance(n p)，物理链路e P的属性有链路带宽bw(e P)、链路时延delay(e P)； The virtual network mapping algorithm based on delay optimization according to claim 1, characterized in that: the virtual network mapping model in step 1 is: the physical network is represented by an undirected weighted graph G P = ( NP , E P ), N P and E P are the physical node set and the physical link set respectively. Among them, the attributes of the physical node n P include CPU resource request cpu(n P ), node delay delay(n P ), and physical node importance importance(n p ), the attributes of the physical link e P include link bandwidth bw (e P ) and link delay delay (e P );
    虚拟网络请求用无向有权图G V＝(N V,E V)表示，N V和E V分别为虚拟节点和虚拟链路集合，其中，虚拟节点n v的资源请求为cpu(n V)、时延属性为delay(n V)，虚拟链路e v的带宽约束为bw(e V)、时延约束为delay(e V)、虚拟节点重要度为importance(n V)。 The virtual network request is represented by an undirected weighted graph G V =(N V , EV ), where NV and EV are virtual nodes and virtual link sets respectively, where the resource request of virtual node n v is cpu(n V ), the delay attribute is delay(n V ), the bandwidth constraint of virtual link ev is bw(e V ), the delay constraint is delay(e V ), and the virtual node importance is importance(n V ).
  3. 根据权利要求1所述基于时延优化的虚拟网络映射算法，其特征在于：步骤2中，虚拟节点n V的重要度计算如下 The virtual network mapping algorithm based on delay optimization according to claim 1, characterized in that: in step 2, the importance of virtual node n V is calculated as follows
    Figure PCTCN2022106693-appb-100001
    Figure PCTCN2022106693-appb-100001
    式中，Res(n V)、Closeness(n V)、VertexDelay(n V)分别为虚拟节点n V的资源集中程度、节点亲密度、节点传播时延期望； In the formula, Res(n V ), Closeness(n V ), and VertexDelay(n V ) are the resource concentration degree, node intimacy, and node propagation delay expectation of virtual node n V respectively;
    物理节点n P的重要度计算如下 The importance of physical node n P is calculated as follows
    Figure PCTCN2022106693-appb-100002
    Figure PCTCN2022106693-appb-100002
    式中，Res(n P)、Closeness(n P)、LocDis(n P)、VertexDelay(n P)分别表示物理节点n P的资源集中程度、节点亲密度、局部网络拓扑重要性、节点传播时延期望。 In the formula, Res(n P ), Closeness(n P ), LocDis(n P ), and VertexDelay(n P ) respectively represent the resource concentration degree, node intimacy, local network topology importance, and node propagation time of the physical node n P. Delay expectations.
  4. 根据权利要求1所述基于时延优化的虚拟网络映射算法，其特征在于：步骤3具体操作如下：The virtual network mapping algorithm based on delay optimization according to claim 1, characterized in that: the specific operation of step 3 is as follows:
    步骤3.1，将虚拟网络请求中的虚拟节点从大到小的排序记录到 VirtualNodeList中；Step 3.1, sort the virtual nodes in the virtual network request from large to small and record them in VirtualNodeList;
    步骤3.2，对于VirtualNodeList中的虚拟节点n V，遍历底层物理节点集合，选取满足物理距离约束条件的候选节点，存入集合Candidates(n V)； Step 3.2, for the virtual node n V in the VirtualNodeList, traverse the underlying physical node set, select the candidate nodes that meet the physical distance constraints, and store them in the set Candidates(n V );
    步骤3.3，判断Candidates(n V)集合是否为空；若为空，此次映射失败，拒绝此次映射并返回映射失败的结果；若不为空，选择Candidates(n V)集合中节点重要度最高的节点，将虚拟节点n V映射至该候选节点上，更新映射关系列表MappingNodeList和底层物理资源，在VirtualNodeList中删除虚拟节点n V； Step 3.3, determine whether the Candidates (n V ) set is empty; if it is empty, this mapping fails, reject this mapping and return the result of mapping failure; if not empty, select the node importance in the Candidates (n V ) set The highest node maps the virtual node n V to the candidate node, updates the mapping relationship list MappingNodeList and the underlying physical resources, and deletes the virtual node n V in the VirtualNodeList;
    步骤3.4，重复步骤3.2和步骤3.3，直至VirtualNodeList为空。Step 3.4, repeat steps 3.2 and 3.3 until VirtualNodeList is empty.
  5. 根据权利要求1所述基于时延优化的虚拟网络映射算法，其特征在于：The virtual network mapping algorithm based on delay optimization according to claim 1, characterized in that:
    步骤4虚拟链路映射采用K-Shortest路径算法，选取最短路径并满足带宽约束，具体操作如下：Step 4: Virtual link mapping uses the K-Shortest path algorithm to select the shortest path and satisfy the bandwidth constraints. The specific operations are as follows:
    步骤4.1，将虚拟网络请求中的虚拟链路从大到小的排序记录到VirtualLinkList中；Step 4.1, sort the virtual links in the virtual network request from large to small and record them in VirtualLinkList;
    步骤4.2，对于VirtualLinkList中的e V虚拟链路，通过K-Shortest路径算法寻找出K条候选物理路径，记作集合Paths(e V)； Step 4.2, for the e V virtual link in VirtualLinkList, K candidate physical paths are found through the K-Shortest path algorithm, which is recorded as the set Paths (e V );
    步骤4.3，对Paths(e V)中的路径进行虚拟链路带宽需求判断，若不能满足虚拟链路带宽需求，则从Paths(e V)中删除该路径； Step 4.3, determine the virtual link bandwidth requirement for the path in Paths(e V ). If the virtual link bandwidth requirement cannot be met, delete the path from Paths(e V );
    步骤4.4，判断Paths(e V)是否为空，若为空，映射失败，返回结果；若不为空，将当前虚拟网络的虚拟链路e V映射至路径优先度最高的物理路径，将映射关系记录进集合MappingLinkList，从VirtualLinkList中删除虚拟链路e V，并更新底层物理资源； Step 4.4, determine whether Paths (e V ) is empty. If it is empty, the mapping fails and the result is returned. If it is not empty, map the virtual link e V of the current virtual network to the physical path with the highest path priority, and map The relationship record is entered into the collection MappingLinkList, the virtual link e V is deleted from the VirtualLinkList, and the underlying physical resources are updated;
    步骤4.5，重复步骤4.2至步骤4.4，直至VirtualLinkList为空。Step 4.5, repeat steps 4.2 to 4.4 until VirtualLinkList is empty.
  6. 根据权利要求5所述基于时延优化的虚拟网络映射算法，其特征在于：步骤4具体操作如下：步骤4.4中路径优先度的计算为The virtual network mapping algorithm based on delay optimization according to claim 5, characterized in that: the specific operation of step 4 is as follows: the calculation of path priority in step 4.4 is:
    Figure PCTCN2022106693-appb-100003
    Figure PCTCN2022106693-appb-100003
    式中，bw(p)为物理路径p的链路带宽，γ为权重因子，在本发明中γ取1，hops(p)为物理路径p的时延，Paths(e V)为虚拟链路e V映射后的候选底层物理路径集合，p为集合Paths(e V)中的一条路径。 In the formula, bw(p) is the link bandwidth of the physical path p, γ is the weight factor, in the present invention, γ is 1, hops(p) is the delay of the physical path p, and Paths(e V ) is the virtual link The set of candidate underlying physical paths after e V mapping, p is a path in the set Paths(e V ).
PCT/CN2022/106693 2022-03-30 2022-07-20 Virtual network mapping algorithm based on delay optimization WO2023184793A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2023507645A JP2024515910A (en) 2022-03-30 2022-07-20 A Delay Optimization-Based Virtual Network Mapping Algorithm

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210322530.2A CN114827273B (en) 2022-03-30 2022-03-30 Virtual network mapping method based on time delay optimization
CN202210322530.2 2022-03-30

Publications (1)

Publication Number Publication Date
WO2023184793A1 true WO2023184793A1 (en) 2023-10-05

Family

ID=82533696

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/106693 WO2023184793A1 (en) 2022-03-30 2022-07-20 Virtual network mapping algorithm based on delay optimization

Country Status (3)

Country Link
JP (1) JP2024515910A (en)
CN (1) CN114827273B (en)
WO (1) WO2023184793A1 (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115767325A (en) * 2022-10-31 2023-03-07 苏州大学 Service function chain profit maximization mapping method and system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018134911A1 (en) * 2017-01-18 2018-07-26 Nec Corporation Resource allocation system, method, and program
CN108667657A (en) * 2018-04-28 2018-10-16 西安交通大学 A kind of mapping method of virtual network based on local feature information towards SDN
CN109150627A (en) * 2018-10-09 2019-01-04 南京邮电大学 The construction method mapped based on dynamic resource demand and the virtual network of topology ambiguity
CN111385151A (en) * 2020-03-11 2020-07-07 中国电子科技集团公司第五十四研究所 Multi-objective optimization-based virtual network mapping method and device

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108900618A (en) * 2018-07-04 2018-11-27 重庆邮电大学 Content buffering method in a kind of information centre's network virtualization
CN111130858B (en) * 2019-12-09 2023-05-19 网络通信与安全紫金山实验室 Dynamic multi-target virtual network mapping method in SD-WAN scene

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018134911A1 (en) * 2017-01-18 2018-07-26 Nec Corporation Resource allocation system, method, and program
CN108667657A (en) * 2018-04-28 2018-10-16 西安交通大学 A kind of mapping method of virtual network based on local feature information towards SDN
CN109150627A (en) * 2018-10-09 2019-01-04 南京邮电大学 The construction method mapped based on dynamic resource demand and the virtual network of topology ambiguity
CN111385151A (en) * 2020-03-11 2020-07-07 中国电子科技集团公司第五十四研究所 Multi-objective optimization-based virtual network mapping method and device

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"Doctoral Dissertation", 16 December 2020, NANJING UNIVERSITY OF POSTS AND TELECOMMUNICATIONS, CN, article CAO, HAOTONG: "Research on Virtual Network Mapping Algorithm in Network Virtualization Environment", pages: 1 - 129, XP009549216, DOI: 10.27251/d.cnki.gnjdc.2020.000704 *

Also Published As

Publication number Publication date
JP2024515910A (en) 2024-04-11
CN114827273A (en) 2022-07-29
CN114827273B (en) 2023-11-24

Similar Documents

Publication Publication Date Title
CN109150627B (en) Virtual network mapping construction method based on dynamic resource demand and topology perception
CN110365514B (en) SDN multistage virtual network mapping method and device based on reinforcement learning
US9633041B2 (en) File block placement in a distributed file system network
US8087025B1 (en) Workload placement among resource-on-demand systems
WO2018120990A1 (en) Method and device for service deployment
CN112738820A (en) Dynamic deployment method and device of service function chain and computer equipment
CN107979545B (en) Virtual network mapping method based on node attributes
CN108616394B (en) Virtual network function backup and deployment method
WO2023184793A1 (en) Virtual network mapping algorithm based on delay optimization
CN110995619B (en) Service quality aware virtual network mapping method and device
CN104917659B (en) A kind of mapping method of virtual network based on virtual network connection performance
Ghosh et al. A game theory-based pricing strategy to support single/multiclass job allocation schemes for bandwidth-constrained distributed computing systems
US7756021B2 (en) Method for finding minimal cost paths under uncertainty
CN108512772A (en) Quality-of-service based data center's traffic scheduling method
CN113300861B (en) Network slice configuration method, device and storage medium
CN111913800B (en) Resource allocation method for optimizing cost of micro-service in cloud based on L-ACO
Quang et al. Evolutionary actor-multi-critic model for VNF-FG embedding
CN115065601A (en) Virtual network mapping method for node link simultaneous mapping
Elzohairy et al. Fedlesscan: Mitigating stragglers in serverless federated learning
CN117749697A (en) Cloud network fusion pre-scheduling method, device and system and storage medium
Shefu et al. Fruit fly optimization algorithm for network-aware web service composition in the cloud
CN116782298A (en) Method for realizing resource preemptive unloading in cloud edge system
Elkael et al. Improved monte carlo tree search for virtual network embedding
CN105917621A (en) Methods and systems for data routing
CN110213149B (en) Virtual network mapping method based on node entropy

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 2023507645

Country of ref document: JP

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22934615

Country of ref document: EP

Kind code of ref document: A1