Cloud service migration stochastic programming method and device aiming at online mobile requirements

Abstract

The invention relates to the technical field of information, in particular to a stochastic programming method and device for cloud service migration aiming at an online mobile demand, and the method comprises the steps: migrating a group of virtual servers through a dynamic migration algorithm based on a local search technology, and dynamically optimizing the migration cost and the access cost; in the service migration process, a migration cost set is given to the node, the migration cost from the current node to the target server is reflected, and a matching function is determined for each server to be used for calculating the migrated target server so as to achieve the minimum access cost. Aiming at the defects of a random planning method for cloud service migration of an online mobile demand, the server can be effectively migrated at the minimum service cost. According to the method, the defects of online and offline algorithms are overcome, and the probability distribution of control data is known or can be estimated. According to the method, the service migration problem which ignores the data segmentation problem can be described by adopting the method with the multi-stage tracing right.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a random planning method and device for cloud service migration for online mobile needs. Background Art

[0002] In the past decade, cloud technology has achieved rapid development. In order to facilitate different end users to access it, it centralizes computing and data resources. Dynamic migration of services can optimize resources according to the current machine load. This technology is currently widely used. On the other hand, with the help of virtualization technology in cloud computing, the requested service is encapsulated in a virtual machine and migrated on demand in the same or across different data centers. This is also becoming a promising way.

[0003] Cloud computing for the mobile world is developing into a widely accepted technology and providing a new generation of services for mobile users. However, the inherent characteristics of mobile access, such as very large scale, temporal and spatial variation, and high sensitivity to service delays, generally make it difficult to implement these services using traditional technologies. In order to cope with these inherent characteristics, it is necessary to migrate services to some advantageous locations in the network close to users, thereby reducing access delays and further reducing the network costs of service providers. Although migration brings many benefits, the cost of migration can also lead to a large amount of data transmission and service interruption. Therefore, how to minimize the cost while still obtaining the benefits of migration is a problem that needs to be solved.

[0004] Here are three sets of related solutions:

[0005] First, Bienkowski et al. proposed a random online migration algorithm for a single server in an n-node network, which explored the competition ratio in low-dynamic request systems and high-dynamic request systems.

[0006] Second, Oikomomou et al. proposed a scalable autonomous network service migration algorithm that observes the differential demand traffic on each link between the node hosting the service and its opposite neighboring nodes. The algorithm proposed by Pantazopoulos et al. only targets a single server and lacks the concept of migration cost.

[0007] Third, Phan et al. proposed a framework called Green Monster to address the energy efficiency problem in cloud computing by utilizing dynamic service migration between Internet data centers (IDCs).

[0008] The above three solutions currently proposed still have some disadvantages.

[0009] In the first solution mentioned above, although they advocate a competitive analysis of the worst-case scenario of the algorithm, they do not conduct a competitive analysis of its general performance.

[0010] In the second solution mentioned above, although this algorithm has certain advantages in service migration, its single-hop migration efficiency is low, so its convergence speed is slow and the network cost is high. In the second solution mentioned above, since it only targets a single server, it does not take into account the situation of multiple servers, and also lacks the concept of migration cost.

[0011] In the third solution mentioned above, the goal is to strike a balance between conflicting optimization goals, and there is no specific goal to minimize the total monetary cost of service migration.

[0012] At the same time, none of the above three solutions were studied directly in a cloud environment.

[0013] In summary, first, in the online approach, service requests are provided to the algorithm one by one in chronological order, and the algorithm makes a migration decision immediately after each request is served, without resorting to any available entire input request from the beginning, and is able to adapt to the access pattern, but because the algorithm does not utilize some statistical information of the input data, the time complexity of each decision is relatively high. Secondly, the offline peer gets the entire request sequence from the beginning, and needs to output a series of migration strategies as a whole to solve the problem at hand, which has the potential to produce high-quality results. However, this is not always feasible in reality, because the entire request sequence is usually not available. Finally, stochastic programming is a method for modeling uncertain optimization problems. Summary of the invention

[0014] The embodiments of the present invention provide a random planning method and device for cloud service migration for online mobile needs, so as to at least overcome the limitation of local search in reducing costs.

[0015] According to an embodiment of the present invention, a random planning method for cloud service migration for online mobile needs is provided, comprising the following steps:

[0016] S101: Migrate a group of virtual servers through a dynamic migration algorithm based on local search technology, and dynamically optimize their migration cost and access cost;

[0017] S102: In service migration, a node is assigned a migration cost set, which reflects the migration cost from the current node to the target server, and a matching function is determined for each server to calculate the migration target server to minimize the access cost.

[0018] Furthermore, the probability distribution of the control data is known or can be estimated. In service migration, when When , a set of k servers are virtually moved using historical access information, where Δ max is the maximum node degree of an n-node network.

[0019] Furthermore, a stochastic programming method is used in dynamic service migration to optimize it so that the expected value of the total cost within the planning range is minimized.

[0020] Furthermore, in any n-node network G(V,E) as a service infrastructure, the service consists of k ≥ 1 servers and runs on virtual machines;

[0021] A collection of host machines Represents a configuration, through batch processing requests σ=σ1σ2...σ m To access; by triggering Server migration in fulfillment requests, a subset of server locations will change frequently over time;

[0022] The subset at time t i The configuration is In t i Batch request at time σ i Using a set v i =∪ j {(a ij ,σ ij )} means that a ij and σ ij is i A subrequest of ij Send to the network.

[0023] Furthermore, batch processing requests σ i The total cost of a visit is calculated as the total connection cost plus the total cost of all requests along all routing paths σ i The transmission cost is expressed as

[0024] in a r is the nearest connection point for request r and Φ(r) is the service node of r determined by the routing function Φ(.), μ is the connection cost, C uv represents the transmission cost between node u and v.

[0025] Furthermore, the migration cost of the service mig Depending on the service size and available bandwidth on the migration path, node v is given a migration cost set To reflect the migration cost of v to the corresponding target server u, middle represents the neighbor set of v.

[0026] Further, a matching function π is determined as Each server in The migration target server in get where π is {1,2,3,...,k}; Indicates from Migrate to The minimum cost.

[0027] Furthermore, for a batch request sequence σ=σ1σ2...σ m , the goal of service migration is to determine To minimize the total service cost, Down, The recursive expression satisfies σ i The total cost of satisfying the first i-1 requests plus the cost of configuring The access cost in arrive migration costs.

[0028] Furthermore, using multi-stage recourse, assuming that each access point a i The number of requests for ∈A is considered as a random process ξ in T time periods i (t);

[0029] Using stochastic programming, assuming ξ i (t) follows a Poisson distribution with a mean of λ i (t), there is only one server in the network G(V, E), where V represents the computing node and E represents the communication link.

[0030] According to another embodiment of the present invention, a random planning device for cloud service migration for online mobile needs is provided, comprising:

[0031] A migration unit, used to migrate a group of virtual servers through a dynamic migration algorithm based on a local search technology, and dynamically optimize the migration cost and access cost thereof;

[0032] The computing unit is used to assign a migration cost set to the node during service migration, reflecting the migration cost from the current node to the target server, and to determine a matching function for each server to calculate the target server for migration to minimize the access cost.

[0033] A storage medium stores a program file capable of implementing any of the above-mentioned random planning methods for cloud service migration for online mobile needs.

[0034] A processor is used to run a program, wherein when the program is run, any of the above-mentioned random planning methods for cloud service migration for online mobile needs is executed.

[0035] The randomized planning method and device for cloud service migration for online mobile demand in the embodiment of the present invention explores these shortcomings and solves them, thereby being able to effectively migrate servers with minimal service cost. It overcomes the shortcomings of online and offline algorithms and utilizes the fact that the probability distribution of control data is known or can be estimated. The present invention can use this method with multi-stage recourse to describe the service migration problem that ignores the data segmentation problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary 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 drawings:

[0037] Figure 1 A flow chart of a random planning method for cloud service migration for online mobile needs according to the present invention;

[0038] Figure 2 A diagram of how mobile edge computing (MEC) servers are deployed;

[0039] Figure 3 This is a module diagram of a random planning device for cloud service migration for online mobile needs according to the present invention. DETAILED DESCRIPTION

[0040] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0041] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0042] Example 1

[0043] According to an embodiment of the present invention, a random planning method for cloud service migration for online mobile needs is provided. Figure 1 , including the following steps:

[0044] S101: Migrate a group of virtual servers through a dynamic migration algorithm based on local search technology, and dynamically optimize their migration cost and access cost;

[0045] S102: In service migration, a node is assigned a migration cost set, which reflects the migration cost from the current node to the target server, and a matching function is determined for each server to calculate the migration target server to minimize the access cost.

[0046] The random programming method for cloud service migration for online mobile demand in the embodiment of the present invention explores these shortcomings and solves them, thereby being able to effectively migrate servers with minimal service cost. It overcomes the shortcomings of online and offline algorithms and utilizes the fact that the probability distribution of control data is known or can be estimated. The present invention can use this method with multi-stage recourse to describe the service migration problem that ignores the data segmentation problem.

[0047] The proposed random programming method for cloud service migration for online mobile demand explores these shortcomings and solves them, thereby enabling efficient server migration with minimal service cost. It overcomes the shortcomings of online and offline algorithms and takes advantage of the fact that the probability distribution of control data is known or can be estimated. The present invention can use this method with multi-stage recourse to describe the service migration problem that ignores the data segmentation problem.

[0048] To this end, this paper proposes a random planning method for cloud service migration for online mobile needs, which can effectively migrate a group of virtual servers at the minimum service cost to adapt to the changes in access patterns. More specifically, when the upper limit of k is , effectively utilize historical access information to perform overall virtual migration on a group of k servers, where Δ max is the maximum node degree of an n-node network, which to some extent overcomes the limitation of local search in reducing costs. Figure 2 This is a typical way to deploy a mobile edge computing (MEC) server. It mainly associates a server with the base station to facilitate the processing of urgent and large-scale processing tasks in the base station.

[0049] The basic contents of the technical solution of the present invention are as follows:

[0050] The present invention proposes a random planning method for cloud service migration targeting online mobile needs.

[0051] First, in any n-node network G(V,E) as a service infrastructure, the service consists of k ≥ 1 servers running on virtual machines. The set of host machines has a configuration, It means that this group of machines processes a series of batch requests σ=σ1σ2...σ from mobile terminals. m To access. By triggering The server migration in satisfies the request, a subset of server locations will change frequently over time. The present invention represents this subset at time t i The configuration is In t i Batch request at time σ i We can use a set σ i =∪ j {(a ij ,σ ij )} means that a ij and σ ij is i A subrequest of ij Send to the network.

[0052] Second, batch processing requests i The total cost of a visit can be calculated as the total connection cost plus the total cost of all requests along all routing paths σ i The transmission cost is expressed as in a r is the nearest connection point for request r and Φ(r) is the service node of r determined by the routing function Φ(.), μ is the connection cost, C uvrepresents the transmission cost between node u and v. The migration cost of the service is Cost mig It depends largely on the service size and available bandwidth on the migration path. The present invention assigns a migration cost set to node v: (Obviously, β vu = 0) to reflect the migration cost of v to the corresponding target server u, middle represents the neighbor set of v.

[0053] Third, in order to minimize the access cost, we need to determine a matching function π, which can be Each server in The present invention indicates that get where π is {1,2,3,...,k}.

[0054] Indicates from Migrate to The minimum cost.

[0055] Fourth, for a batch request sequence σ=σ1σ2...σ m , the goal of service migration is to determine (t is certain) so that the total service cost is minimized, that is, Down,

[0056] This recursive expression satisfies σ i The total cost of satisfying the first i-1 requests plus the cost of configuring The access cost in arrive Through this method, the present invention not only migrates the service to a favorable location in the network close to the user, reducing the access delay, but also reduces the network cost of providing the service.

[0057] The technical solution of the present invention is described in detail as follows:

[0058] The present invention proposes a random planning method for cloud service migration for online mobile needs, which uses a multi-stage recourse, ignoring the data segmentation problem involved and assuming that each access point a i The number of requests for ∈A is considered as a random process ξ in T time periods i (t). The method proposed in the present invention is suitable for random programming, because ξ iThe probability distribution of (t) can be estimated from historical data. On the other hand, migration decisions are made repeatedly under essentially the same circumstances, and the goal is to come up with a decision that performs well on average. In particular, the present invention assumes in the formula that ξ i (t) follows a Poisson distribution with a mean of λ i (t), and are independent in the formula stage; for simplicity, there is only one server in the network G(V, E), and the node operation cost is not considered, where V represents the computing node and E represents the communication link.

[0059] First, assuming that the initial t = 1, the server is located at u1, in each cycle t = 1, 2, ..., T, the algorithm first observes the current server location u t , then migrate to position v t , which leads to the migration cost Cost mig (u t ,v t ). After the server migration, implement random request ξ i (t), 1≤i≤|A|, denoted as That is, the position of the next server, that is, at the beginning of cycle t+1, u t+1 =v t . Then the total cost incurred in period t is:

[0060]

[0061] Cost mig (u t ,v t ) represents the migration cost of the server from node u to node v in time period t, which can be expressed as β·SP(u t ,v t ) definition representation.

[0062] Cost acc (ξ i (t),x t ) represents ξ in time period t i (t) The access cost of requesting service x can be expressed as ξ i (t)(W μ +L λ SP(a i ,x t )) definition represents. (SP(x,y) represents the shortest path from x to y in the network G).

[0063] Secondly, this method is to minimize the expected value of the total cost within the planning range. Therefore, the present invention proposes the following optimization method:

[0064]

[0065] Because u t+1 =v t , where t=1, ..., T-1, so the present invention can understand formula (2) as a dynamic optimization process.

[0066] Consider the requesting process R t , where t = 1, ..., T-1. The present invention uses R [t] =(R1,...,R t ) is the history of the request progress until time t, where And through Special Implementation

[0067] Afterwards, due to unexpected constraints, the present invention calculates the target position v within the time period t. t The decision should depend only on the information available at the time of decision making, which is the observations made during the request. [t-1] , so in the final stage when t=T, the problem that the present invention needs to solve is:

[0068]

[0069] Obviously, the optimal value (or optimal solution set) of problem (3) depends on u T and r [T-1] Looking back in time, the present invention has the following dynamic programming equation:

[0070]

[0071] Where t=T-1,...,2. Finally, in the first stage, the present invention needs to solve this problem:

[0072]

[0073] Because i (t) is stage independent, so ξ i (t) and R [t-1] The conditional expectations in equations (3)-(5) become the corresponding unconditional expectations, so F t (u t ) does not depend on the implementation of the request, but only becomes the respective single variable u t It can be rewritten as an equivalent deterministic dynamic programming problem (t = T-1,...,1):

[0074]

[0075] where u t+1 =vt And d(a i ,x)=L λ SP(a i ,x t ). As the computation progresses over time, the bounds are

[0076]

[0077] let is the optimal solution of equation (6) under the assumption of stage independence. Become u t Separate function. However, t Itself is and decision (v1,...,v t-1 ). Therefore, v t =v t (r [t-1] ) is a function of the request flow realizations available at decision time. Where t=1,...,T-1.

[0078] The key points and protection points of the present invention are:

[0079] First, it uses local search technology to develop an efficient multi-server migration method. This method is different from previous methods. In service migration, When , historical access information is effectively used to perform virtual movement of a set of k servers.

[0080] Secondly, the migration cost of the service depends largely on the service size and available bandwidth on the migration path, so this method will assign a migration cost set to the node to reflect the migration cost from the current node to the target server; in order to minimize the access cost, a matching function will be determined for each server to calculate the target server for migration.

[0081] Finally, this method uses the random programming method in dynamic service migration. In order to minimize the expected value of the total cost within the planning range, it is optimized, that is,

[0082]

[0083] Compared with the prior art, the advantages of the present invention are:

[0084] 1. First, by migrating a group of virtual servers through this dynamic migration algorithm based on local search technology and dynamically optimizing its migration cost and access cost, the total service cost of the request demand can be further reduced. It is precisely because stochastic programming has a strong applicability for solving uncertain optimization problems that it is used in this method. It can overcome the shortcomings of online and offline algorithms by taking advantage of the fact that the probability distribution of control data is known or can be estimated.

[0085] 2. Secondly, our approach outperforms the local search algorithm in various situations including those without virtual mobility, and is extremely cost-effective in service migration while having a faster convergence speed.

[0086] 3. After that, it is difficult for general distributed algorithms to achieve global load balancing, but the method in this article can effectively achieve global load balancing between servers.

[0087] 4. Finally, the complexity of server location combinations can be overcome, further ensuring that k servers are eventually migrated to a set of suboptimal service nodes and remain there if the request pattern does not change significantly. This helps a group of servers move in the network to provide services for user requests.

[0088] Example 2

[0089] According to another embodiment of the present invention, a random planning device for cloud service migration for online mobile needs is provided. Figure 3 ,include:

[0090] A migration unit 201 is used to migrate a group of virtual servers through a dynamic migration algorithm based on a local search technology, and dynamically optimize the migration cost and access cost thereof;

[0091] The calculation unit 202 is used to assign a migration cost set to a node during service migration, reflecting the migration cost from the current node to the target server, and to determine a matching function for each server to calculate the migration target server to minimize the access cost.

[0092] The random planning device for cloud service migration for online mobile demand in the embodiment of the present invention and the random planning method for cloud service migration for online mobile demand explore these shortcomings and solve them, thereby being able to effectively migrate servers with minimal service cost. It overcomes the shortcomings of online and offline algorithms and utilizes the fact that the probability distribution of control data is known or can be estimated. The present invention can use this method with multi-stage recourse to describe the service migration problem that ignores the data segmentation problem.

[0093] The proposed random programming method for cloud service migration for online mobile demand explores these shortcomings and solves them, thereby enabling efficient server migration with minimal service cost. It overcomes the shortcomings of online and offline algorithms and takes advantage of the fact that the probability distribution of control data is known or can be estimated. The present invention can use this method with multi-stage recourse to describe the service migration problem that ignores the data segmentation problem.

[0094] To this end, the present invention proposes a random planning device for cloud service migration for online mobile needs, which can effectively migrate a group of virtual servers at the minimum service cost to adapt to the change of access mode. More specifically, when the upper limit of k is , effectively utilize historical access information to perform overall virtual migration on a group of k servers, where Δ max is the maximum node degree of an n-node network, which to some extent overcomes the limitation of local search in reducing costs. Figure 2 This is a typical way to deploy a mobile edge computing (MEC) server. It mainly associates a server with the base station to facilitate the processing of urgent and large-scale processing tasks in the base station.

[0095] The basic contents of the technical solution of the present invention are as follows:

[0096] The present invention proposes a random planning device for cloud service migration targeting online mobile needs.

[0097] First, in any n-node network G(V,E) as a service infrastructure, the service consists of k ≥ 1 servers running on virtual machines. The set of host machines has a configuration, It means that this group of machines processes a series of batch requests σ=σ1σ2...σ from mobile terminals. m By triggering server migration in H to satisfy the request, a subset of server locations will change frequently over time. The present invention represents this subset at time t i The configuration is In t i Batch request at time σ i We can use a set σ i =∪ j {(a ij ,σ ij )} means that a ij and σ ij is i A subrequest of ij Send to the network.

[0098] Second, batch processing requests iThe total cost of a visit can be calculated as the total connection cost plus the total cost of all requests along all routing paths σ i The transmission cost is expressed as in a r is the nearest connection point for request r and Φ(r) is the service node of r determined by the routing function Φ(.), μ is the connection cost, C uv represents the transmission cost between node u and v. The migration cost of the service is Cost mig It depends largely on the service size and available bandwidth on the migration path. The present invention assigns a migration cost set to node v: (Obviously, β vu = 0) to reflect the migration cost of v to the corresponding target server u, middle represents the neighbor set of v.

[0099] Third, in order to minimize the access cost, we need to determine a matching function π, which can be Each server in The present invention indicates that get where π is {1,2,3,...,k}.

[0100] Indicates from Migrate to The minimum cost.

[0101] Fourth, for a batch request sequence σ=σ1σ2...σ m , the goal of service migration is to determine (t is certain) so that the total service cost is minimized, that is, Down,

[0102] This recursive expression satisfies σ i The total cost of satisfying the first i-1 requests plus the cost of configuring The access cost in arrive Through this method, the present invention not only migrates the service to a favorable location in the network close to the user, reducing the access delay, but also reduces the network cost of providing the service.

[0103] The technical solution of the present invention is described in detail as follows:

[0104] The present invention proposes a random planning device for cloud service migration for online mobile needs, which uses a multi-stage recourse, ignoring the data segmentation problem involved, and assuming that each access point a i The number of requests for ∈A is considered as a random process ξ in T time periods i (t). The method proposed in the present invention is suitable for random programming, because ξ i The probability distribution of (t) can be estimated from historical data. On the other hand, migration decisions are made repeatedly under essentially the same circumstances, and the goal is to come up with a decision that performs well on average. In particular, the present invention assumes in the formula that ξ i (t) follows a Poisson distribution with a mean of λ i (t), and are independent in the formula stage; for simplicity, there is only one server in the network G(V, E), and the node operation cost is not considered, where V represents the computing node and E represents the communication link.

[0105] First, assuming that the initial t = 1, the server is located at u1, in each cycle t = 1, 2, ..., T, the algorithm first observes the current server location u t , then migrate to position v t , which leads to the migration cost Cost mig (u t ,v t ). After the server migration, implement random request ξ i (t), 1≤i≤|A|, denoted as That is, the position of the next server, that is, at the beginning of cycle t+1, u t+1 =v t . Then the total cost incurred in period t is:

[0106]

[0107] Cost mig (u t ,v t ) represents the migration cost of the server from node u to node v in time period t, which can be expressed as β·SP(u t ,v t ) definition representation.

[0108] Cost acc (ξ i (t),x t ) represents ξ in time period t i (t) The access cost of requesting service x can be expressed as ξ i (t)(W μ +L λ SP(a i ,xt )) definition represents. (SP(x,y) represents the shortest path from x to y in the network G).

[0109] Secondly, this method is to minimize the expected value of the total cost within the planning range. Therefore, the present invention proposes the following optimization method:

[0110]

[0111] Because u t+1 =v t , where t=1, ..., T-1, so the present invention can understand formula (2) as a dynamic optimization process.

[0112] Consider the requesting process R t , where t = 1, ..., T-1. The present invention uses R [t] =(R1,...,R t ) is the history of the request progress until time t, where And through Special Implementation

[0113] Afterwards, due to unexpected constraints, the present invention calculates the target position v within the time period t. t The decision should depend only on the information available at the time of decision making, which is the observations made during the request. [t-1] , so in the final stage when t=T, the problem that the present invention needs to solve is:

[0114]

[0115] Obviously, the optimal value (or optimal solution set) of problem (3) depends on u T and r [T-1] Looking back in time, the present invention has the following dynamic programming equation:

[0116]

[0117] Where t=T-1,...,2. Finally, in the first stage, the present invention needs to solve this problem:

[0118]

[0119] Because i (t) is stage independent, so ξ i (t) and R [t-1] The conditional expectations in equations (3)-(5) become the corresponding unconditional expectations, so F t (u t) does not depend on the implementation of the request, but only becomes the respective single variable u t It can be rewritten as an equivalent deterministic dynamic programming problem (t = T-1,...,1):

[0120]

[0121] where u t+1 =v t And d(a i ,x)=L λ SP(a i ,x t ). As the computation progresses over time, the bounds are

[0122]

[0123] let is the optimal solution of equation (6) under the assumption of stage independence. Become u t Separate function. However, t Itself is and decision (v1,...,v t-1 ). Therefore, v t =v t (r [t-1] ) is a function of the request flow realizations available at decision time. Where t=1,...,T-1.

[0124] The key points and protection points of the present invention are:

[0125] First, it uses local search technology to develop an efficient multi-server migration method. This method is different from previous methods. In service migration, When , historical access information is effectively used to perform virtual movement of a set of k servers.

[0126] Secondly, the migration cost of the service depends largely on the service size and available bandwidth on the migration path, so this method will assign a migration cost set to the node to reflect the migration cost from the current node to the target server; in order to minimize the access cost, a matching function will be determined for each server to calculate the target server for migration.

[0127] Finally, this method uses the random programming method in dynamic service migration. In order to minimize the expected value of the total cost within the planning range, it is optimized, that is,

[0128]

[0129] Compared with the prior art, the advantages of the present invention are:

[0130] 1. First, by migrating a group of virtual servers through this dynamic migration algorithm based on local search technology and dynamically optimizing its migration cost and access cost, the total service cost of the request demand can be further reduced. It is precisely because stochastic programming has a strong applicability for solving uncertain optimization problems that it is used in this method. It can overcome the shortcomings of online and offline algorithms by taking advantage of the fact that the probability distribution of control data is known or can be estimated.

[0131] 2. Secondly, our approach outperforms the local search algorithm in various situations including those without virtual mobility, and is extremely cost-effective in service migration while having a faster convergence speed.

[0132] 3. After that, it is difficult for general distributed algorithms to achieve global load balancing, but the method in this article can effectively achieve global load balancing between servers.

[0133] 4. Finally, the complexity of server location combinations can be overcome, further ensuring that k servers are eventually migrated to a set of suboptimal service nodes and remain there if the request pattern does not change significantly. This helps a group of servers move in the network to provide services for user requests.

[0134] Example 3

[0135] A storage medium stores a program file capable of implementing any of the above-mentioned random planning methods for cloud service migration for online mobile needs.

[0136] Example 4

[0137] A processor is used to run a program, wherein when the program is run, any of the above-mentioned random planning methods for cloud service migration for online mobile needs is executed.

[0138] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0139] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0140] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the system embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0141] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0142] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0143] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0144] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A random planning method for cloud service migration for online mobile needs, characterized in that: The following steps are involved: S101: Migrate a group of virtual servers through a dynamic migration algorithm based on local search technology, and dynamically optimize their migration cost and access cost; S102: In service migration, a node is assigned a migration cost set, which reflects the migration cost from the current node to the target server, and a matching function is determined for each server to calculate the migration target server to minimize the access cost.

2. The random planning method for cloud service migration for online mobile needs according to claim 1, characterized in that: The probability distribution of control data is known or can be estimated. In service migration, when When , a set of k servers are virtually moved using historical access information, where Δ max is the maximum node degree of an n-node network.

3. The random planning method for cloud service migration for online mobile needs according to claim 1, characterized in that: The random programming method is used in dynamic service migration to optimize it so that the expected value of the total cost within the planning range is minimized.

4. The random planning method for cloud service migration for online mobile needs according to claim 2, characterized in that: In any n-node network G(V,E) as a service infrastructure, the service consists of k ≥ 1 servers and runs on virtual machines; A collection of host machines Represents a configuration, through batch processing requests σ=σ1σ2...σ m To access; by triggering Server migration in fulfillment requests, a subset of server locations will change frequently over time; The subset at time t i The configuration is In t i Batch request at time σ i Use a set v i =∪ j {(a ij ,σ ij )} means that a ij and σ ij is i A subrequest of ij Send to the network.

5. The random planning method for cloud service migration for online mobile needs according to claim 4, characterized in that: Batch Requestsv i The total cost of a visit is calculated as the total connection cost plus the total cost of all requests v along all routing paths i The transmission cost is expressed as in a r is the nearest connection point for request r and Φ(r) is the service node of r determined by the routing function Φ(.), μ is the connection cost, C uv represents the transmission cost between node u and v.

6. The random planning method for cloud service migration for online mobile needs according to claim 5, characterized in that: Cost of service migration mig Depending on the service size and available bandwidth on the migration path, node v is given a migration cost set To reflect the migration cost of v to the corresponding target server u, middle represents the neighbor set of v.

7. The random planning method for cloud service migration for online mobile needs according to claim 6, characterized in that: Determine a matching function π, as Each server in The migration target server in get where π is {1,2,3,...,k}; Indicates from Migrate to The minimum cost.

8. The random planning method for cloud service migration for online mobile needs according to claim 7, characterized in that: For a batch request sequence σ=σ1σ2...σ m , the goal of service migration is to determine To minimize the total service cost, Down, The recursive expression satisfies σ i The total cost of satisfying the first i-1 requests plus the cost of configuring The access cost in arrive migration costs.

9. The random planning method for cloud service migration for online mobile needs according to claim 8, characterized in that: Using multi-stage recourse, assuming that each access point a i The number of requests for ∈A is considered as a random process ξ in T time periods i (t); Using stochastic programming, assuming ξ i (t) follows a Poisson distribution with a mean of λ i (t), there is only one server in the network G(V, E), where V represents the computing node and E represents the communication link.

10. A random planning device for cloud service migration for online mobile needs, characterized in that: include: A migration unit, used to migrate a group of virtual servers through a dynamic migration algorithm based on a local search technology, and dynamically optimize the migration cost and access cost thereof; The computing unit is used to assign a migration cost set to a node during service migration, reflecting the migration cost from the current node to the target server, and to determine a matching function for each server to calculate the target server for migration to minimize the access cost.

Images