CN111698703B - Network reliability optimization method based on business priority and load balancing - Google Patents

Abstract

The present invention provides a network reliability optimization method based on business priority and load balancing, which is applied to a virtualized wireless sensor network. The step of updating the resource allocation mapping of the virtual network in the virtualized wireless sensor network includes: in the first At the first moment, mark the virtual network that needs to be migrated, and calculate the load balancing rate of the physical network layer at this moment; at the second moment, for a virtual network that needs to be migrated, if its new resource allocation mapping is executed, the load balancing rate of the physical network layer will be If the load balancing rate is less than the load balancing rate at the first moment, a new resource allocation mapping of the virtual network is performed. The present invention optimizes network reliability based on theoretical analysis of business priority and load balancing, thereby improving the service quality of the virtualized wireless sensor network and extending the service time.

Description

Network reliability optimization method based on business priority and load balancing

Technical field

The invention relates to the field of resource management of wireless sensor networks, and in particular to a virtualized network reliability optimization method based on business priority and load balancing.

Background technique

With the rapid development and application of Internet of Things technology (IoT), sensor networks (Sensor Networks) represented by Wireless Sensor Networks (WSN) have been applied to various fields such as industry, medical care, transportation, and environment. Sensor network is a network formed by free organization and combination of sensor nodes with computing, communication and sensing capabilities. Since wireless sensor networks are widely distributed, the power supply of sensor nodes is generally provided by their own power supplies.

In order to maximize the resource utilization of sensor nodes and ensure the normal operation of services on the sensor network, resource allocation and reliability optimization of wireless sensor networks have become a research focus. Because virtualization technology can effectively improve the utilization of network resources and the reliability of network services, wireless virtualized sensor network technology (VirtualizedWireless Sensor Networks) based on virtualization has been proposed and has been widely used. In the virtualized wireless sensor network environment, the existing sensor network is divided into a physical network layer and a virtual network layer. The physical network nodes of the physical network layer include physical sensors and provide physical network resources for the virtual network of the virtual network layer. Each virtual network in the virtual network layer rents the resources of the physical network layer according to business needs, thereby realizing various business operation services on the virtual network. Regarding the resource allocation and network reliability optimization of sensor networks, existing research can be divided into two types: traditional network environment and network virtualization environment.

In traditional network environments, two strategies are mainly used: routing protocol optimization and rapid fault recovery. In terms of routing protocol optimization, the literature [Ren Xiuli, Chen Yang. Routing protocol for data transmission delay optimization in wireless sensor networks [J]. Computer Applications, 2019, 40(1):196-201.] to solve the problem of wireless transmission With the goal of reliable transmission of data packets in the sensor network, the data transmission efficiency is analyzed based on network detection technology, so as to optimize the routing of transmission links with high packet loss rates. Literature [Wang Junhong, Chen Yuedong, Chen Mengyuan. Research on real-time and reliability optimization of smart distribution network WSNs based on fuzzy cognitive graph [J]. Journal of Sensing Technology, 2016, 29(2): 213-219.] For smart grid To solve the problem of high failure rate, the reliability of wireless sensor network can be optimized by adjusting routing paths and network parameters. Literature [Wang Guan, Wang Ruiyao. Self-powered wireless sensor network routing algorithm based on cluster head optimization [J]. Computer Applications, 2018, 6(1721): 1725-1736.] to solve the problem of low data transmission success rate in wireless sensor networks Taking the problem as the goal, the cluster head selection mechanism in the grouping algorithm is optimized to ensure the reliability of the sensor network in each group.

In a network virtualization environment, there are two main strategies: optimal resource allocation to improve utilization and resource allocation considering survivability. Literature [Delgado C, Canales M, Ortín J, et al.Joint applicationadmission control and network slicing in virtual sensor networks [J]. IEEEInternet of Things Journal, 2017, 5(1): 28-43.] uses an exhaustive search method Solve the optimal business scheduling sequence to improve the service quality of the sensor network. Literature[Soualah O,Aitsaadi N,Fajjari I.A novelreactive survivable virtual network embedding scheme based on game theory[J].IEEE Transactions on Network and Service Management,2017,14(3):569-585.] to solve the problem of virtual sensor network In order to solve the problem of low efficiency in access control, network slicing and admission control technology are combined to improve the flexibility of sensor network access control. Regarding the virtual resource mapping algorithm that considers survivability, literature [10] adopts link redundancy to solve the link failure problem. Literature [Shahriar N, Chowdhury S R, Ahmed R, etal.Virtual network survivability through joint spare capacity allocation and embedding[J]. IEEE Journal on Selected Areas in Communications, 2018,36(3):502-518.] Combine capacity backup with The virtual network embedding process is merged, thereby effectively improving the reliability of the network.

There have been studies on optimizing resources during the resource allocation process and in the event of failures. When allocating resources, reserved resources are used, which can easily lead to resource waste. Optimizing resources in the event of a failure affects the service quality of the virtual network. In addition, due to the imbalance of resource allocation, it is easy for some physical network nodes to exhaust their energy by carrying virtual network services for a long time, causing the physical network layer to become unavailable. Therefore, it is necessary to balance the allocation of resources carrying virtual services on physical network nodes to ensure stable operation of the physical network layer.

Contents of the invention

The purpose of the present invention is to provide an optimization method for a virtualized wireless sensor network, which optimizes network reliability based on theoretical analysis of business priority and load balancing, thereby improving the service quality of the virtualized wireless sensor network and extending the service time.

The technical solution provided by the present invention is a network reliability optimization method based on business priority and load balancing. The steps of updating the resource allocation mapping of the virtual network in the virtualized wireless sensor network include: at the first moment, marking the need for migration virtual network, and calculate the load balancing rate of the physical network layer at this moment; at the second moment, for a virtual network that needs to be migrated, if its new resource allocation mapping is executed, the load balancing rate of the physical network layer is less than the first If the load balancing rate at the time is determined, the new resource allocation mapping of the virtual network will be performed.

Preferably, the method of marking the virtual network that needs to be migrated is: at a moment, select a physical network node whose remaining life cycle is less than a threshold in the physical network layer, and mark the physical network node corresponding to each virtual service it carries. The virtual network is the virtual network that needs to be migrated.

A further improvement is that for multiple virtual networks that need to be migrated, the virtual network with the lowest delay requirement is given priority to provide new resource allocation mapping.

A further improvement is that for a virtual network that needs to be migrated, migration is attempted after obtaining a new resource allocation mapping.

A further improvement is that for a virtual network that needs to be migrated, its new resource allocation mapping is obtained by solving the shortest path at the physical network layer. At the same time, for the new resource allocation mapping, the delay between the corresponding virtual network nodes can be judged by the hop count of the path between the physical network nodes, so as to determine whether the new resource allocation mapping meets the delay requirements of the virtual network.

A further improvement is that for a virtual network that needs to be migrated, if the path hop count of its new resource allocation mapping is greater than the delay requirement of the virtual network, its new resource allocation mapping will not be performed.

A further improvement is that the step of updating the resource allocation mapping of the virtual network in the virtualized wireless sensor network includes:

S100, at a moment, obtain the physical network nodes whose remaining life cycle is less than a threshold in the physical network layer of the virtualized wireless sensor network, and mark the virtual networks carried by these physical network nodes as virtual networks that need to be migrated; obtain the The load balancing coefficient of the physical network layer at any time

S200: Traverse all virtual networks that need to be migrated, starting from the virtual network with the lowest delay requirement, and allocate to it a new resource allocation mapping obtained through the shortest path algorithm; if the new resource allocation mapping satisfies the requirements of the virtual network If the delay requirements and the load balancing requirements of the physical network layer are met, the virtual network will be migrated according to the new resource allocation mapping. Otherwise, the virtual network will not be migrated;

S300: Repeat steps S100 and S200 and count it as one optimization. If the number of repetitions exceeds the preset optimization number threshold, or there is no virtual network that needs to be migrated, the optimization process ends.

A further improvement is that step S100 includes the following steps:

S101, for any physical network node n _i ∈N, calculate its remaining life cycle

S102, the remaining life cycle Less than life cycle threshold/> The physical network nodes are put into the set Θ;

S103. Put the virtual network corresponding to the virtual service carried on each physical network node in the set Θ into the virtual network set Ω, and mark it as a virtual network that needs to be migrated;

S104, calculate the current load balancing coefficient of the physical network layer and assign it to

A further improvement is that step S200 includes the following steps:

S201. Traverse each virtual network in the set Ω and calculate its delay requirements. That is, for any virtual network in the set Ω There is a delay requirement H _x ;

S202: Arrange each virtual network in the set Ω in ascending order according to the value of the delay requirement of each virtual network to obtain the set Ω _ord ;

S203, select a virtual network in the set Ω _ord Initialize the loop variable x=1, that is, when executing this step for the first time, you should select a virtual network with the smallest delay requirement in the set Ω _ord . Before each subsequent execution of this step, x is incremented by one;

S204, for the current virtual network For virtual services, use the Dijkstra algorithm to solve the shortest path to obtain a new virtual network that assumes the current virtual network/> The topology structure of the physical network layer of the virtual service, and calculate the number of path hops of the topology/>

S205, determine the current virtual network Whether it is less than its delay requirement H _x . If greater, execute step S203;

S206: Calculate the load balancing coefficient η _reload of the physical network layer under the new topology of the virtual network, and determine whether it satisfies If satisfied, jump to step S208;

S207. If there are other paths solved by the Dijkstra algorithm in step 204, select one of the sub-shortest paths as the current virtual network. Map the topology structure and return to step S205;

S208. Execute migration and transfer the selected virtual network Map to the path of the current topology structure and return to step S203.

A further improvement is that step S300 includes the following steps:

S301, determine whether the optimization times threshold T is exceeded. If it has been exceeded, end;

S302. For each physical node n _i ∈N, evaluate whether its remaining life cycle is less than the threshold. If satisfied, end; if not satisfied, execute step S102.

In order to achieve load balancing of wireless sensor network resources and ensure that the life cycle of physical nodes is as long as possible, thereby ensuring the service quality of the virtual network, the present invention proposes a network reliability optimization method based on business priority and load balancing. algorithm based on service priority and load balancing in wireless sensor network (NROA-SPoLB)). The algorithm includes three steps: S100 to select the virtual network that needs to be migrated, S200 to migrate the virtual network, and S300 to evaluate whether the life cycle of all physical nodes meets the threshold. In step S100, by selecting physical nodes that need to be optimized, all virtual services of physical nodes that do not meet the life cycle are migrated. The advantage of this is to reserve more space for physical network nodes to prevent their resources from being exhausted when re-mapping. In step S200, for the virtual services that need to be migrated, the shorter their delay, the more they need to be migrated first, so as to obtain shorter links and meet the delay requirements. In step S300, an optimization times threshold T is set to limit the number of execution times of the algorithm to prevent over-calculation, which may lead to excessive computing resource overhead. Secondly, determine whether the life cycle of all physical nodes n _i ∈N meets the threshold It can effectively prevent some physical node resources from being overused due to migration.

Description of the drawings

The description and drawings that constitute a part of this application are used to provide a further understanding of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the attached picture:

Figure 1 is a schematic flow chart of a network reliability optimization method based on business priority and load balancing according to the present invention, and

Figure 2 is a schematic diagram of the impact of the number of virtual networks on the proportion of effective nodes in an embodiment including a comparison method, and

Figure 3 is a schematic diagram of the impact of physical network node scale on the proportion of effective nodes in an embodiment including a comparison method.

Detailed ways

It should be noted that, as long as there is no conflict, the embodiments and features in the embodiments of this application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

In the solution of the present invention, in order to solve the problem of low reliability of wireless sensor equipment, the present invention provides a network reliability optimization method based on business priority and load balancing.

In order to achieve load balancing of wireless sensor network resources and ensure that the life cycle of physical nodes is as long as possible, thereby ensuring the service quality of the virtual network, the present invention proposes a network reliability optimization method based on business priority and load balancing. The algorithm based on service priority and load balancing in wireless sensor networks (NROA-SPoLB) is shown in Figure 1. The method includes steps S100 to S300:

S100, select the virtual network that needs to be migrated;

S200, migrate virtual networks in sequence;

S300, if there is a virtual network that needs to be migrated, repeat S100 and S200, otherwise end.

In step S100, by selecting physical nodes that need to be optimized, all virtual services of physical nodes that do not meet the life cycle are migrated. The advantage of this is to reserve more space for physical network nodes to prevent their resources from being exhausted when re-mapping.

In step S200, for the virtual services that need to be migrated, the shorter their delay, the more they need to be migrated first, so as to obtain shorter links and meet the delay requirements.

In step S300, an optimization times threshold T is set to limit the number of execution times of the algorithm to prevent over-calculation, which may lead to excessive computing resource overhead. Secondly, determine whether the life cycle of all physical nodes n _i ∈N meets the threshold It can effectively prevent some physical node resources from being overused due to migration.

The methods of various embodiments of the present invention are based on a network virtualization environment. In the network virtualization environment, the wireless sensor network is virtualized and divided into a physical network layer and a virtual network layer to form a virtualized wireless sensor network. The virtual network layer Provide multiple virtual networks to provide services to multiple virtual services at the same time. Chinese patent CN110933728A discloses a mapping device for a virtualized wireless sensor network. Its control layer is equivalent to the virtual network layer of the present invention, and its sensing infrastructure layer is equivalent to the physical network layer of the present invention. An embodiment of the method of the present invention. Can run in this public control layer controller. An embodiment of the present invention running on a cloud computing resource management platform belongs to the configuration method of Resource Orchestration Service (ROS), wherein:

The structural topology of the physical network layer is represented by an undirected graph G(N,L). Among them, N represents the set of physical network nodes, and E represents the set of physical network links. For any physical network node n _i ∈N, it at least includes CPU computing power C( _ni ), node position P(xi _, y _i ), residual energy Three self-attributes, the subscript i represents the index number of the physical network node. L is the set of physical network links between physical network nodes. For a physical network link l _ij ∈L between a physical network node n _i and a physical network node n _j , it at least includes the bandwidth B(l _ij ). Self-property, among which, i≠j, l _ij =l _ji .

The structural topology of a virtual network provided by the virtual network layer is represented by an undirected graph G ^v (N ^v , L ^v ). Among them, N ^v represents the set of virtual network nodes, and L ^v represents the set of virtual network links. For a virtual network node The at least one attribute it includes is CPU computing power/> For a virtual network node/> with virtual network nodes/> virtual network link/> Its own attributes include at least bandwidth/>

In various embodiments of the present invention, for a virtual network request pointing to a specific virtual service, as a response, a virtual network is generated at the virtual network layer to provide services for the virtual service, and the virtual network has a delay limit condition, to describe the quality of its business services. In one embodiment of the present invention, M virtual networks are generated in response to M virtual network requests, using Represents a set of M virtual networks. There are K delay constraints corresponding to each virtual network in the set. The set of K delay constraints is expressed as/> and Obviously, M≥K. For virtual network migration, that is, as a response to the same virtual network request, under an external condition, a virtual network/> Its actual resource allocation mapping in the physical network layer has changed. Therefore, without changing the topology of the physical network layer, the attributes of the physical network nodes or physical network links in the topology have changed.

In various embodiments of the present invention, as a response to a virtual network request, when allocating physical network layer resources to a virtual network, that is, allocating CPU resources of physical network nodes to several of its virtual network nodes, and/or, When allocating the bandwidth resources of physical network links to several virtual network links, for a physical network node n _i , use C ^init (n _i ) to represent its initial CPU calculation before that moment, that is, before the allocation is completed. Capability, use C ^used (n _i ) to represent its used CPU computing power after this moment, that is, after the allocation is completed. For a physical network link l _ij , use B ^init (l _ij ) to represent its usage at this moment. The initial bandwidth before, that is, before the allocation is completed, uses B ^left (l _ij ) to represent the remaining bandwidth after this moment, that is, after the allocation is completed.

In various embodiments of the present invention, the remaining energy of a physical network node is used as a criterion to measure its life cycle. The larger the remaining life cycle, the better the reliability of the physical network node in undertaking virtual network services. In one embodiment of the present invention, formula (1) is used to describe the remaining life cycle of the physical network node n _i ∈ N at a specific time t

in, Indicates the total energy consumption of the physical network node before time t,/> Indicates the remaining energy of the physical network node at time t. It can be seen from formula (1) that the remaining life cycle of a physical network node at a moment is inversely proportional to its total energy consumption and directly proportional to its remaining energy. /> The value is the sum of the historical energy consumption of sending data E _send and the energy consumption of receiving data E _receive before time t.

Exemplarily, the energy consumption E _send of a physical network node sending k-bit data is calculated using formula (2).

The energy consumption E _receive for receiving k-bit data is calculated using formula (3).

_Ereceive =kE ₀ (3)

Among them, E ₀ represents the radio frequency energy consumption coefficient, which is related to the physical attributes of the physical network node itself; pre represents the indicator vector, which is used to adjust the exponential growth relationship between the topological distance d _ij and energy consumption. For a physical network node, There is a threshold d ₀ . The topological distance d _ij has a specified index pre when it is less than the threshold, and has a larger index pre when it is greater than the threshold. For example, when d _ij <d ₀ , the preferred pre=2 ; When d _ij ≥ d ₀ , preferred pre=4. d ₀ is the threshold of d _ij , preferably, ε _fs represents the energy attenuation coefficient, and ε represents the multipath attenuation coefficient. d _ij represents the topological distance between the physical network node n _i ∈N and its next-hop physical network node n _j ∈N. When the distance is preferably Euclidean distance, it can be calculated using formula (4).

Based on the above exemplary description, it can be obtained A calculation process of can be shown as formula (5).

One aspect of the concept of the present invention is to ensure that each physical network node has a large life cycle at all times, and each service needs to be evenly distributed on the physical network nodes to avoid excessive energy consumption of some physical network nodes. And premature disability. In one embodiment of the present invention, η _reload is used to represent the load balancing rate of a physical network layer at a time t, and is calculated using formula (6).

in, Respectively represent the maximum and minimum values of the remaining energy in all its physical network nodes. It can be seen from formula (6) that the load balancing rate η _reload of the physical network layer at a specific moment is a number greater than 1. The closer the value is to 1, the more balanced the remaining energy of all physical network nodes is. This type of formula can be used Determine whether the energy consumption of each physical network node is balanced. Generally, the load balancing rate η _reload of the physical network layer is basically stable under the same resource allocation strategy. Similarly, use/> Indicates the load balancing rate of the physical network layer after the last resource allocation policy was implemented. When/> When , it indicates that the current resource allocation strategy can better meet the load balancing of network resources at the physical network layer.

One aspect of the concept of the present invention is that the communication time from one virtual network node to another virtual network node in the virtual network x, or the time delay _T The attributes of the physical network layer are associated, and the present invention converts the delay T _x between virtual network nodes into the number of hops H _x between its mapped physical network nodes. In one embodiment of the invention, a virtual network The number of hops H _x is calculated using formula (7).

in, Indicates the average data processing time and transmission time of each virtual network link in virtual network x.

Exemplarily, based on the above embodiments, in a complete embodiment, steps S100 to S300 of the method of the present invention are implemented through the following specific steps.

This embodiment is implemented by a resource orchestration module running on a cloud computing platform. The cloud computing platform connects and allocates resources of physical network nodes of a wireless sensor network and provides virtualized wireless sensor network services to multiple businesses. Among them, Its resource orchestration module adjusts the topology structure of the physical network layer of the virtualized wireless sensor network according to virtual network requests to ensure the effective node proportion of each physical network node in the physical network layer.

In this embodiment, based on response or API call, the cloud computing platform provides the resource orchestration module with a topology structure G(N,E) of the physical network layer. In this topology state, all services actually carried by the physical network layer are Corresponding to M virtual networks, the set of all virtual networks is The resource orchestration module provides the cloud computing platform with the new resource allocation strategy G′(N′,E′), G′(N′,E′) and G(N) under the topological structure of the physical network layer through the method of the present invention. ,E) The topology is the same, but the attributes of the physical network nodes and physical network links are different, so that the cloud computing platform can adjust the physical network layer to provide updated virtualized wireless sensor network services.

Specifically, the resource orchestration module is configured to provide a new topology G′(N′,E′) according to the following steps.

S100: According to a topology structure G(N,E) of the physical network layer, extract the virtual network that needs to be migrated from all the virtual networks carried by the topology structure.

Exemplarily, this step includes the following sub-steps S101 to S104. in,

S101, for any physical network node n _i ∈N, use formula (1) to calculate its remaining life cycle

S102, set the remaining life cycle to be less than the life cycle threshold The physical network nodes are put into the set Θ; for different physical network nodes, there can be different life cycle thresholds; the same physical network node can have different life cycle thresholds at different times.

S103. Put the virtual network corresponding to the services carried on each physical network node in the set Θ into the virtual network set Ω, and mark it as a virtual network that needs to be migrated;

S104, use formula (6) to calculate the current load balancing coefficient of the physical network layer and assign it to

S200, migrate virtual network.

Exemplarily, this step includes the following sub-steps S201 to S208. in,

S201, traverse each virtual network in the set Ω, and use formula (7) to calculate its delay requirement, that is, for any virtual network in the set Ω There is a delay requirement H _x ;

S202: Arrange each virtual network in the set Ω in ascending order according to the delay requirement of each virtual network to obtain the set Ω _ord ;

S203, select a virtual network in the set Ω _ord Initialize the loop variable x=1, that is, when executing this step for the first time, the virtual network with the smallest delay requirement in the set Ω _ord should be selected. Each time this step is executed in the future, x will be incremented by one;

S204, for the current virtual network Use Dijkstra's algorithm to find the shortest path to obtain a new virtual network that assumes the current virtual network/> A topology structure of the physical network layer of the virtual service, and calculate the number of path hops of the topology/> The topology structure includes a number of physical network nodes and a number of physical network links that bear the virtual service;

S205: Determine whether the current virtual network is under the selected path. Whether it is less than its delay requirement H _x . If it is not satisfied, the migration fails and the migration is not performed, and the process goes to step S203;

S206: Calculate the load balancing coefficient η _reload of the physical network layer under the new topology of the virtual network, that is, the load balancing coefficient under the new resource allocation mapping of the virtual network, and determine whether it satisfies If satisfied, jump to step S208;

S207. If there are other paths solved by the Dijkstra algorithm in step 204, select one of the sub-shortest paths as the current virtual network. Map the topology structure and return to step S205;

S208. Execute migration and transfer the selected virtual network Map to the path of the current topology structure and return to step S203; the virtual network is considered to be a delay-sensitive virtual network.

S300: Evaluate whether the life cycle of all physical nodes meets the threshold; repeat steps S100 and S200 and count it as an optimization. If the number of repetitions exceeds the preset optimization number threshold, or there is no virtual network that needs to be migrated, then the optimization The process ends.

Exemplarily, this step includes the following sub-steps S301 to S302. in,

S301, determine whether the optimization times threshold T is exceeded. If it has been exceeded, end;

S302. For each physical node n _i ∈N, evaluate whether the life cycle meets the threshold. If satisfied, end; if not satisfied, put into the set Θ to be optimized. If the set Θ is not empty, return to step S102.

Using the above method embodiments, the present invention provides the following different specific environments as multiple specific embodiments to further illustrate the effects of the present invention.

In each specific embodiment, considering that the maximizing recovery of available nodes Algorithm (MRANA) is a relatively typical network reliability optimization method, the algorithm of the present invention is compared with the algorithm MRANA. Among them, the MRANA algorithm selects all unavailable nodes and uses the shortest path method to remap the virtual network on them. In terms of evaluation indicators, the proportion of effective nodes in the physical network layer is used for evaluation. The proportion of effective physical network nodes refers to the number of physical network nodes with remaining life cycles greater than the threshold, as a proportion of the total number of physical network nodes.

Specific embodiment one

In this embodiment, in the virtual network layer, the number of virtual networks is increased from 50 to 100 to simulate the impact of the number of virtual network requests of different sizes on algorithm performance. For each virtual network, the number of virtual network nodes obeys the uniform distribution of (2, 4), and the CPU resources of each virtual network node obeys the uniform distribution of (1, 5). The bandwidth resources of each virtual network link obey the uniform distribution of (1, 3). In terms of parameter settings, the remaining life cycle threshold of the physical network node is set to 10%, the value of ε is set to 100, the value of ε _fs is set to 10, the value of E ₀ is set to 50, and the initial energy of the node is set to 100 , the transmission energy consumption of 1 bit data is 0.02.

In this embodiment, the number of physical network nodes is maintained at 200, and the comparison results of the impact of the number of virtual networks on effective nodes are shown in Figure 2. In the figure, the X-axis represents the increase in the number of virtual networks from 50 to 100. It can be seen from the figure that as the number of virtual networks increases, the proportion of effective nodes under both algorithms increases, indicating that both algorithms can optimize more physical network nodes as the number of virtual networks increases. Comparing the two algorithms, the network optimization capability of the algorithm of the present invention is stronger, indicating that the algorithm of the present invention can better optimize network resources.

Specific embodiment two

In this embodiment, the number of physical network nodes in the physical network layer in the wireless sensor network increases from 100 to 500, and the CPU resources of each physical network node obey the uniform distribution of (40, 50). The bandwidth resources of each physical network link obey the uniform distribution of (20, 30).

In this embodiment, the number of virtual networks remains at 70, and the comparison results of the impact of physical network scale on effective nodes are shown in Figure 3. In the figure, the X-axis represents the network topology when the number of physical network nodes increases from 100 to 500. It can be seen from the figure that as the number of physical network nodes increases, the proportion of effective nodes under both algorithms increases, indicating that the number of physical network nodes increases, and the physical network resources that can be selected by the virtual network increase rapidly, thereby better realizing the network Resource optimization. Comparing the two algorithms, the algorithm of the present invention has stronger network optimization capabilities.

Claims (1)

1. A network reliability optimization method based on business priority and load balancing, characterized in that the step of updating the resource allocation mapping of the virtual network in the virtualized wireless sensor network includes: at the first moment, selecting the virtual For physical network nodes whose remaining life cycle is less than a threshold in the physical network layer of the wireless sensor network, mark the virtual network corresponding to each virtual service carried by the physical network node as a virtual network that needs to be migrated, and calculate the load of the physical network layer at this moment. Balancing rate; at the second moment, for a virtual network that needs to be migrated, if the load balancing rate of the physical network layer is less than the load balancing rate of the first moment after its new resource allocation mapping is executed, the new virtual network will be executed. Resource allocation mapping; in, For a virtual network that needs to be migrated, try to perform migration after obtaining the new resource allocation mapping; For a virtual network that needs to be migrated, its new resource allocation mapping is obtained by solving the shortest path at the physical network layer; For a virtual network that needs to be migrated, if the path hop count of its new resource allocation mapping is greater than the delay requirement of the virtual network, its new resource allocation mapping will not be executed; The step of updating the resource allocation mapping of the virtual network in the virtualized wireless sensor network includes: S100, at a moment, obtain the physical network nodes whose remaining life cycle is less than a threshold in the physical network layer of the virtualized wireless sensor network, and mark the virtual networks carried by these physical network nodes as virtual networks that need to be migrated; obtain the The load balancing coefficient of the physical network layer at any time S200: Traverse all virtual networks that need to be migrated, starting from the virtual network with the lowest delay requirement, and allocate to it a new resource allocation mapping obtained through the shortest path algorithm; if the new resource allocation mapping satisfies the requirements of the virtual network If the delay requirements and the load balancing requirements of the physical network layer are met, the virtual network will be migrated according to the new resource allocation mapping. Otherwise, the virtual network will not be migrated; S300: Repeat steps S100 and S200, and count it as one optimization. If the number of repetitions exceeds the preset optimization number threshold, or there is no virtual network that needs to be migrated, then the optimization process ends; The step S100 includes the following steps: S101, for any physical network node n _i ∈N, calculate its remaining life cycle S102, the remaining life cycle Less than life cycle threshold/> The physical network nodes are put into the set Θ; S103. Put the virtual network corresponding to the virtual service carried on each physical network node in the set Θ into the virtual network set Ω, and mark it as a virtual network that needs to be migrated; S104, calculate the current load balancing coefficient of the physical network layer and assign it to The step S200 includes the following steps: S201. Traverse each virtual network in the set Ω and calculate its delay requirements. That is, for any virtual network in the set Ω There is a delay requirement H _x ; S202: Arrange each virtual network in the set Ω in ascending order according to the value of the delay requirement of each virtual network to obtain the set Ω _ord ; S203, select a virtual network in the set Ω _ord Initialize the loop variable x=1, that is, when executing this step for the first time, you should select a virtual network with the smallest delay requirement in the set Ω _ord . Before each subsequent execution of this step, x is incremented by one; S204, for the current virtual network For virtual services, use the Dijkstra algorithm to solve the shortest path to obtain a new virtual network that assumes the current virtual network/> The topology structure of the physical network layer of the virtual service, and calculate the number of path hops of the topology/> S205, determine the current virtual network Whether it is less than its delay requirement H _x ; if it is greater, execute step S203; S206: Calculate the load balancing coefficient η _reload of the physical network layer under the new topology of the virtual network, and determine whether it satisfies If satisfied, jump to step S208; S207. If there are other paths in the Dijkstra algorithm solution in step 204, select one of the sub-shortest paths as the current virtual network. Map the topology structure and return to step S205; S208, perform migration and transfer the selected virtual network Map to the path of the current topology structure and return to step S203; The step S300 includes the following steps: S301, determine whether the optimization times threshold T is exceeded. If it has been exceeded, end; S302. For each physical network node n _i ∈N, evaluate whether its remaining life cycle is less than the threshold. If all are satisfied, end; if there are physical network nodes that are not satisfied, execute step S102.