CN102662764B - A kind of dynamic cloud computational resource optimizing distribution method based on SMDP - Google Patents

Abstract

The invention discloses a kind of dynamic cloud computational resource optimizing distribution method based on SMDP, belong to computer communication technology field.This method is: 1) user satisfaction is divided into N class by cloud computing service domain system; 2) terminal user sends services request to cloud computing service territory, and application uses cloud computing service; 3) cloud computing service domain system sets up an action collection according to the services request received and current cloud computing service territory state; 4) for each action in action collection, the long-term gain in cloud computing service territory is calculated; 5) cloud computing service domain system determines whether to accept current service request according to the long-term gain calculated, if accepted, the VM Resource Allocation Formula choosing the maximum action correspondence of long-term gain is cloud computing service request dispatching VM.Compared with prior art, the present invention substantially increases user satisfaction and the service quality of mobile terminal.

Description

A kind of dynamic cloud computational resource optimizing distribution method based on SMDP

Technical field

The invention belongs to computer communication technology field, relate to the resource optimal distribution method of cloud computing system, particularly relate to the Optimal Configuration Method to the cloud computing resources in cloud computing service territory in mobile cloud computing system.

Background technology

Cloud computing is that one is distributed according to need with resource, pay-as-you-go, usefulness is calculated as the new calculation services pattern (Armbrust of feature, M., Fox, A., Griffith, R., Joseph, A., Katz, R., Konwinski, A., Lee, G., Patterson, D., Rabkin, A., Stoica, I., et al. " Above the clouds:A berkeley view of cloud computing " .EECS Department, University of California, Berkeley, Tech.Rep.UCB/EECS-2009-28 (2009)).Cloud computing is not only cloud computing service business simultaneously for personal user provides a kind of new computation schema yet, it can be divided into infrastructure-as-a-service (IaaS), platform-as-a-service (PaaS) and software-as-a-service (SaaS) three major types by broad sense.Along with the development of wireless communication technology and Internet technology, mobile terminal will replace PC gradually becomes global topmost internet access facility.Because mobile terminal (MD) has more advantage (such as movability compared with traditional wire terminal, dirigibility and perception etc.), therefore the nature that mobile computing and cloud computing technology combined just becomes the new method building Mobile solution, no matter has also attracted increasing concern in academia or industry member at present.Thus, a new research field-mobile cloud computing (Mobile Cloud Computing) is also just arisen at the historic moment.

In the former research about mobile cloud computing, what main research direction concentrated on calculation task uploads download, long-range operation and dynamic organization etc.Author is at (X.Li, H.Z, and Y.Zhang, " Deploying Mobile Computation in Cloud Service " in Proceedings of the First International Conference for Cloud Computing (CloudCom), 2009, p.301.) propose one can be run Mobile solution mobile cloud computing model at mobile terminal and high in the clouds in, thus calculating, transmission and store tasks can be uploaded to high in the clouds operation by the mobile terminal of resource-constrained.Author is at (B.Chun and P.Maniatis, " Augmented Smartphone Applications Through Clone Cloud Execution; " in Proceedings of USENIX HotOS XII, 2009.) in by increase perform number of times configure CloneCloud cloud resource, but do not consider the actual motion state of user terminal.The Resourse Distribute that mobile terminal is served flexible application by system for cloud computing is at (X.Zhang, J.Schiffman, S.Gibbs, A.Kunjithapatham, and S.Jeong, " Securing elastic applications on mobile devices for cloud computing, " in Proceedings of the 2009 ACM workshop on Cloud computing security, 2009, pp.127-134.) the inside done some preliminary researchs.At document (D.Huang, X.Zhang, M.Kang, and J.Luo, " Mobicloud:A secure mobile cloud framework for pervasive mobile computing and communication, " in Proceedings of 5th IEEE International Symposium on Service-Oriented System Engineering, 2010.) inner, the people such as Huang propose mobile cloud computing framework, the virtual machine (VM-Virtual Machine) that this model allows mobile terminal related application to be uploaded to high in the clouds runs.Author is at (X.Meng, V.Pappas, and L.Zhang, " Improving the scalability of data center networks with trafficaware virtual machine placement, " in IEEE INFOCOM, San Diego, CA, USA, March 2010.) in propose a kind of different flow according to different geographical and carry out configuring virtual machine, improved the new method of the utilization factor of network by the placement location of distributing virtual machine rationally.In fact, because these research and inquirement about the framework facility of mobile cloud computing are relatively more abundant, therefore, the Resourse Distribute of mobile cloud computing will become Next main direction of studying naturally.

In mobile system for cloud computing, based on the distributed placement of server group on geographic position, the cloud computing resources (such as CPU, internal memory and storage etc.) of system carrys out responsible distribution by multiple mobile cloud computing service territory respectively.Each moves cloud computing service territory and is made up of multiple virtual machine (VM-Virtual Machine), and each virtual machine (VM) is then made up of the minimum cloud computing resources that can process a cloud computing service.Although compared with mobile terminal, the cloud computing resources of mobile system for cloud computing is considered to unlimited usually, is still necessary very much to make full use of cloud computing resources in mobile cloud computing service territory to realize the low cost movement of mobile system for cloud computing.

Especially the research that the resources configuration optimization moving cloud computing to cloud computing at present carries out is also fewer.Document (H.Liang, D.Huang, and D.Peng, " On Economic Mobile Cloud Computing Model, " in Proceedings of the International Workshop on Mobile Computing and Clouds (MobiCloud in conjunction with MobiCASE), 2010.) an Eco-power mobile cloud computing resources apportion model is proposed, this model can when given system configuration, by optimizing the maximum return of distributing Mobile solution and obtaining mobile system for cloud computing beyond the clouds and between mobile terminal.Document (G.Wei, A.V.Vasilakos, Y.Zheng, and N.Xiong, " A game-theoretic method of fair resource allocation for cloud computing services; " 2009.) propose one based on game theoretic cloud computing resources apportion model, this model can distribute cloud computing resources according to the demand of mobile terminal to QoS of customer (QoS).In addition, also have some documents how to be optimized by the server of virtual machine or data center system for cloud computing to distribute cloud computing resources and be studied.At (K.Lorincz, B.r.Chen, J.Waterman, G.Werner-Allen, and M.Welsh, " Resource aware programming in the pixie os; " in SenSys ' 08, Raleigh, North Carolina, USA, November 2008.) in, author proposes a new cloud computing operation model, and this operation model can not only make user programme when grasping cloud computing resources, also can realize in system for cloud computing, cloud computing service reuses the allocation model of cloud computing resources simultaneously.Document (K.Lorincz, B.Chen, J.Waterman, G.Werner-Allen, and M.Welsh, " A stratified approach for supporting high throughput event processing applications; " in DEB S ' 09, Nashville, TN, USA, July2009.) the cloud computing resources distribution of event application in system for cloud computing is studied.At (G.Tesauro, N.K.Jong, R.Das, and M.N.Bennani, " A hybrid reinforcement learning approach to autonomic resource allocation; " in Proc.of ICAC-06, Dublin, Ireland, June 2006.) in, author proposes a resource allocator model based on enhancement mode self learning system and carries out dynamic assignment to the server in system for cloud computing, thus improves the income of system for cloud computing.At (K.Boloor, R.Chirkova, Y.Viniotis, and T.Salo, " Dynamic request allocation and scheduling for context aware applications subject to a percentile response time sla in a distributed cloud, " in 2 ^ndiEEE International Conference on Cloud Computing Technology and Science, Indianapolis, Indiana, USA, November 2010.) in, author proposes a general scheme of distributing cloud computing service request and planning, the program, while obtaining user's service quality of specifying, improves the income of cloud computing service provider.

Optimization for cloud computing resources distributes, and domesticly it is also proposed some solutions.Such as in patented claim 201110097395.8 (a kind of management controls system for cloud computing technological system), author (Cao Xuezhu) proposes a kind of invention managing control system for cloud computing technological system; In patented claim 201110138021.6 (a kind of cloud computing resource management system and method), author (Ji Xinhua, Nie Song, Du Hai and horse strong) proposes the invention of a kind of cloud computing resource management system and method; In patented claim 201110075410.9 (management method of system for configuration information in cloud computation operation and system), author's (Zhang Liqiang and Haitao Zhang) proposes a kind of management method of system for configuration information in cloud computation operation and the invention of system; In patented claim 201080005003.4 (system and method for automated management of virtual resources in cloud computing environment), author (SM You Mubaihaoke) proposes a kind of invention for the system of managing virtual resource in cloud computing environment; In patented claim 201110222073.1 (a kind of cloud computing management system based on virtual resources), author (Shen Lingyun, Ruan Minhui and Zhou Yongfeng) proposes the invention of a kind of cloud computing management system based on virtual resources (C2MS).A main advantage of mobile system for cloud computing is the Mobile solution service allowing mobile terminal to run them beyond the clouds.And a cloud computing service can also be assigned with the cloud computing resources of multiple VM to make the mobile terminal higher calculating of acquisition and storage capacity.When mobile cloud computing service territory receive one send over from mobile terminal cloud computing service request time, the current available cloud computing resources of system Water demand, and based on analysis result determine whether receive this cloud computing service request; If determine it is receive, so system also needs judgement to be further specially cloud computing service request dispatching how many cloud computing resources (i.e. the number of VM) of this mobile terminal.If cloud computing resources all in mobile cloud computing service territory is occupied, so due to the deficiency of cloud computing resources, system can refuse the cloud computing service request (we suppose in mobile cloud computing, do not have buffer queue) of this mobile terminal.Not only on the user satisfaction of mobile terminal and service quality, negative impact is brought to the refusal of mobile terminal cloud computing service request, and also greatly reduces the net proceeds of system.

The system income in mobile cloud computing service territory increases along with the increase of received cloud computing service number of requests usually.But then, along with the cloud computing service request of system acceptance is more, the cloud computing resources so distributing to each cloud computing service is also fewer, thus reduces the user satisfaction of mobile terminal and the system performance in mobile cloud computing service territory that accept to serve.And the existing income that only considered system about cloud computing resources distribution method major part, do not consider the expenditure that cloud computing resources is occupied brought, do not consider user satisfaction and the service quality (QoS) of mobile terminal yet.Therefore, in order to the comprehensive system benefit in mobile cloud computing service territory can be obtained, when calculating the system benefit in mobile cloud computing service territory, not only need the income considering mobile system for cloud computing, also need user satisfaction and the service quality (QoS) of expenditure and the mobile terminal considering that cloud computing resources is occupied brought.

Summary of the invention

For the Optimizing Allocation of the cloud computing resources in mobile system for cloud computing cloud computing service territory, the object of the present invention is to provide a kind of dynamic cloud computational resource optimization method based on SMDP.Present invention newly proposes the mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation based on semi-morkov decision processes (SMDP), the optimization Decision of Allocation strategy of the cloud computing resources in mobile cloud computing service territory is obtained by this model, and obtain the maximum return in mobile cloud computing service territory, this income not only considers the income receiving cloud computing service request and bring, have also contemplated that because cloud computing service takies the expenditure that cloud computing resources brings simultaneously, and the user satisfaction of mobile terminal and service quality (QoS).Therefore, this invention all has very important effect to the integral benefit of mobile cloud computing system and mobile terminal client to the raising of the satisfaction of mobile system for cloud computing, and this is also actual value place of the present invention.

Technical scheme of the present invention is:

Based on a dynamic cloud computational resource optimizing distribution method of SMDP, the steps include:

1) user satisfaction is divided into N class by cloud computing service domain system, and satisfaction classification is the virtual machine VM number that user's correspondence of i is distributed is k _i; Wherein, 1≤k _i≤ K, K are the VM sum in cloud computing service territory;

2) terminal user sends services request to cloud computing service territory, and application uses cloud computing service;

3) cloud computing service domain system sets up an action collection according to the services request received and current cloud computing service territory state;

4) for each action in described action collection, the long-term gain in cloud computing service territory is calculated;

5) cloud computing service domain system determines whether to accept current service request according to the long-term gain calculated, if accepted, the VM Resource Allocation Formula choosing the maximum action correspondence of long-term gain is cloud computing service request dispatching VM.

Further, the state s in cloud computing service territory is expressed as s=<n ₁, n ₂..., n _n, e>; Wherein, n _ifor satisfaction classification in cloud computing service territory is the number of users of i, e is the event in cloud computing service territory, e ∈ { R, D ₁, D ₂, D _i...., D _n, R is cloud computing service request, D _ithe cloud computing service being i for satisfaction classification completes and releases the VM number shared by it.

Further, described action collection is $A\left(s\right)=\left\{\begin{array}{cc}-1& e\&Element;\{D_1,D_2,...,D_N\}\\ \{\mathrm{0,1},...,N\},& e=R.\end{array}\right.;$ Wherein, A (s)=-1 represents that cloud computing service terminates to run and discharges the state of shared cloud computing resources, A (s)=0 represents the refusal cloud computing service request of cloud computing service territory, and A (s)=i represents that cloud computing service territory receives cloud computing service request and the cloud computing resources distributing to this cloud computing service request is k _iindividual VM; S represents the current state in cloud computing service territory.

Further, utilize formula z (s, a)=x (and s, a)-τ (s, a) y (s, a) calculate for each action a income z (s, a); Wherein, x (s, a) be state be s, the action of selection is when being a, the gross income that cloud computing service territory obtains, τ (s, a) represent state be s, the action chosen be a time, transfer to the service time desired by next state j; Y (s, a) represent state be s, the action chosen be a time cloud computing service territory expenditure.

Further, formula is utilized $x(s,a)=\left\{\begin{array}{cc}-1,& e=R,a=0\\ U\left(k_i\right),& e=R,a=i\end{array}\right.$ Calculate the gross income that cloud computing service territory obtains, U (k _i) be Efficiency Function.

Further, formula is utilized the service time desired by next state is transferred to by current state in calculating cloud computing service territory; Wherein, α is the discount rate between two decision points under continuous time, and decision point refers to the time point that any one event e occurs, τ ₁refer to the time experienced to the state that next event occurs from current state.

Further, time τ between two decision points (s, a) obeys index distribution, and mean speed γ that event occurs (s, a)=τ (s, a) ^-1.

Further, cloud computing service domain system utilizes formula $v\left(s\right)=z(s,a)+\mathrm{\&eta;}\underset{j\&Element;S}{\mathrm{\&Sigma;}}p\left(j\right|s,a)v\left(j\right),$ Calculate the long-term gain v (s) in cloud computing service territory during each action a in described action collection; Wherein, p (j|s, a) be state transition probability, j is the NextState in mobile cloud computing service territory, γ (s, a) be the mean speed that event occurs, α is the discount rate between two decision points under continuous time, and decision point refers to the time point that any one event e occurs, S is all possible state in cloud computing service territory, the overall long-term gain obtained when v (j) represents NextState j.

Further, adopt an Efficiency Function to measure the satisfaction of cloud computing user, user satisfaction is divided into N class.

Compared with prior art, good effect of the present invention is:

The present invention is based on semi-morkov decision processes (SMDP), first proposed a new mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation, the dynamic cloud computational resource that obtained by this model optimizes Decision of Allocation strategy can not only make the system benefit in mobile cloud computing service territory maximum, also can improve the utilization factor of mobile cloud computing service territory cloud computing resources and the user satisfaction of mobile terminal and service quality (QoS) simultaneously.With traditional greedy algorithm to compared with the allocative decision of Internet resources, the optimisation strategy that the mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation proposed according to us obtains, its system benefit and performance have all had and have significantly improved.From Fig. 5 and Fig. 6, along with the growth of cloud computing service request arriving rate, especially when the arrival rate of cloud computing service request is more than 5, efficiency earnings of the present invention improves more than at least 50% (as shown in Figure 5) compared with conventional greedy algorithm, and blocking rate of the present invention then at least reduced for more than 50% (as shown in Figure 6) compared with conventional greedy algorithm simultaneously.

The main contribution of the present invention is embodied in following three aspects:

1) the dynamic cloud computational resource optimization Decision of Allocation strategy in mobile cloud computing service territory is deduced based on semi-morkov decision processes (SMDP).

2) this model can based on the current available cloud computing resources in mobile cloud computing service territory, for cloud computing service request distributes different cloud computing resources adaptively, the cloud computing resources moving cloud computing service territory by making full use of this improves cloud computing resources utilization factor, and obtains the largest global income in mobile cloud computing service territory.

3) the maximum system income in the mobile cloud computing service territory of this model acquisition, both considered this to move cloud computing service territory and receive the income that cloud computing service request brings, have also contemplated that because of occupied the brought expenditure of cloud computing resources, also contemplate user satisfaction and the service quality (QoS) of mobile terminal.Therefore, the system benefit obtained by this model is comprehensive integral benefit.

Mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation proposed by the invention, not only can improve the utilization factor of the cloud computing resources in mobile system for cloud computing cloud computing service territory, also can improve the service quality (QoS) of mobile subscriber simultaneously.In order to verify the performance of mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation proposed by the invention, the performance of itself and conventional greedy algorithm (Greedy Algorithm) is compared (R.Ramjee by by experiment, D.Towsley, and R.Nagarajan, " On optimal call admission control in cellular networks, " Wireless Networks, vol.3, no.1, pp.29-41,1997).Our experimental result shows, the mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation that application the present invention proposes, the entire system income of mobile system for cloud computing improves more than 50% compared with greedy algorithm, the unaccepted probability of its cloud computing service request then decreases more than 50% compared with greedy algorithm, and performance and the service quality (QoS) of mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation also namely proposed by the invention all improve more than 50% compared with greedy algorithm.

Accompanying drawing explanation

Fig. 1 is the service model of mobile system for cloud computing;

Fig. 2 is method flow diagram of the present invention;

Fig. 3 is the Efficiency Function of multimedia service;

Fig. 4 is state transition diagram (N=2), and wherein, on the arrow line of connection two states, Section 1 (such as a=0) represents the action taked under current state, and on the arrow line of connection two states, Section 2 (such as ) represent under current state, after taking corresponding action, transfer to the transition probability of next state;

Fig. 5 efficiency earnings of the present invention and conventional greedy algorithm comparison diagram;

Fig. 6 blocking rate of the present invention and conventional greedy algorithm comparison diagram.

Embodiment

Below in conjunction with accompanying drawing, the present invention is explained in further detail.

1. move the arthmetic statement of cloud computing service territory dynamic cloud computational resource model of optimizing allocation:

The main advantage of mobile system for cloud computing compared with traditional Client-Server service mode is: when mobile terminal their application service uploaded to high in the clouds carry out computing time, mobile terminal can obtain more capacity and better performance (such as less processing time, the saving etc. of battery capacity of mobile terminal).Uploading of mobile terminal flexible application task can be realized by the Weblet connecting high in the clouds and mobile terminal.Weblet can use Java or .Net independent of platform or Python, also can usage platform programming language.At (B.Chun and P.Maniatis, " Augmented Smartphone Applications Through Clone Cloud Execution; " in Proceedings of USENIX HotOS XII, 2009.) have studied in Weblet is uploaded to the algorithm run in high in the clouds from mobile terminal.Upload flexible application by Weblet to serve high in the clouds and run, mobile terminal significantly can improve self computing power, storage capacity and the network bandwidth etc.Usually, mobile terminal determines whether task to be uploaded to high in the clouds and runs the state (such as, the electricity of the CPU processing power of mobile terminal, battery, network connection quality and mobile terminal are to factors such as the considerations of safety) depending on mobile terminal self.In the present invention, when mobile terminal determines that task being uploaded to high in the clouds runs, first it can send a services request to high in the clouds, if high in the clouds receives the services request of mobile terminal, so task will be uploaded to high in the clouds operation by mobile terminal subsequently, after end of run, operation result can be returned to mobile terminal by high in the clouds.

In the present invention, the computational resource in mobile system for cloud computing and the communication resource (comprise the CPU in server, memory device, internal memory etc., and other routing device and communication facilitiess etc.) be all carry out unified management by virtual machine (VM).As shown in Figure 1, a VM is in charge of Weblet uploading, unload and processing in mobile system for cloud computing.As previously mentioned, in mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation proposed by the invention, a VM is the minimum cloud computing resources (CPU of process needed for a cloud computing service in mobile cloud computing service territory, internal memory and storage etc.), the cloud computing resources assigned by each VM once can only process a cloud computing service request.Although we can think that the cloud computing resources in mobile system for cloud computing is unlimited, in mobile system for cloud computing, certain concrete mobile cloud computing service territory is again limited with the cloud computing resources that VM quantity counts.Therefore, in mobile cloud computing service territory, if the quantity of the cloud computing service request arrived has exceeded cloud computing resources VM number available in this service-domain, then the cloud computing service request arrived subsequently will have been refused by this service-domain.On the other hand, if the quantity of the cloud computing service request arrived is far below cloud computing resources VM number available in this service-domain, so this service-domain just can make full use of the cloud computing resources of this service-domain for the more VM number of each cloud computing service request dispatching, improves this move the cloud computing resources utilization factor in cloud computing service territory and the user satisfaction of mobile terminal and service quality (QoS) with this.

Therefore, the target of mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation proposed by the invention is exactly by making full use of cloud computing resources, make mobile cloud computing service territory can obtain largest global system benefit, also can improve the cloud computing resources utilization factor of this service-domain and user satisfaction and service quality (QoS).

In the present invention, we consider a mobile system for cloud computing only comprising a cloud computing service territory, if its cloud computing resources is total up to K virtual machine (VM).Represent the VM number of the cloud computing resources distributing to a mobile terminal with k, wherein k is a positive integer, and the 1≤k that satisfies condition≤K.In addition, we can with different Efficiency Function (J.W.Lee, R.R.Mazumdar, and N.B.Shroff, " Non-convex optimization and rate control for multi-class services in the Internet, " IEEE/ACM Transactions on Networking, 2005, vol.13, no.4, pp.827-840.) measure the satisfaction of mobile cloud computing user.Such as, the user satisfaction of mobile cloud computing mobile terminal can be described with similar Sigmoidal function,

$U\left(r\right)=1-\exp (-\frac{{\mathrm{\&omega;}}_2{r}^2}{{\mathrm{\&omega;}}_1+r}),---\left(1\right)$

Wherein U (r) represents the user satisfaction of mobile system for cloud computing, and r is the cloud computing resources that mobile terminal is distributed in mobile cloud computing service territory, ω ₁and ω ₂be used to the parameter regulating U (r) waveform, the waveform of its function as shown in Figure 3.

Generally, parameter ω ₁and ω ₂selection be that the demand to service quality (QoS) decides by mobile cloud computing service and final user, effective Index selection distributes the cloud computing resources in mobile cloud computing service territory and has significant impact.From formula (1), in order to improve the user satisfaction of mobile system for cloud computing, need to distribute cloud computing resources as much as possible to mobile phone users.But, on the other hand, in order to improve the overall system income of mobile system for cloud computing, move the total available cloud computing resources in cloud computing service territory according to this and mobile subscriber uses cloud computing service to the demand of cloud computing resources, system can not distribute maximum cloud computing resources to separately each mobile phone users again.In order to Optimized model can be set up to the dynamic need of the cloud computing service of mobile system for cloud computing, we suppose that mobile terminal request accesses mobile system for cloud computing and uses the process of cloud computing service to obey Poisson distribution (Poisson), its average is λ, that also supposes mobile cloud computing terminal user is netting time obeys index distribution simultaneously, and its average is λ represents the speed of the cloud computing request reaching system for cloud computing with Poisson distribution, and μ represents that the user terminating service leaves the speed of system for cloud computing obeys index distribution.

2. the step as shown in Figure 2, setting up mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation is as follows:

1) system state is set

In order to semi-morkov decision processes can be used to characterize the model of optimizing allocation of mobile cloud computing service territory cloud computing resources, because user was inversely proportional to the time that the satisfaction of mobile system for cloud computing and its services request are processed, namely the satisfaction of time shorter then user to system that be processed of the services request of user is higher, obviously, if the cloud computing resources (i.e. VM number) distributing to this user is more, the time that then this user service of asking is processed then shorter (as a rule, cloud computing service can carry out parallel processing by the cloud computing resources of multiple VM, thus improve travelling speed).It can thus be appreciated that user is directly proportional to the cloud computing resources distributing to this user to the satisfaction of mobile system for cloud computing, the cloud computing resources namely distributing to user is more, then the satisfaction of this user is higher.The user satisfaction of mobile system for cloud computing is divided into N class by us.Therefore the system state that our definition the present invention is based on the mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation of Semi-Markov process is the cloud computing service quantity that has under each user satisfaction and moves the set of institute's event in cloud computing service territory at this.We also define k _ifor distributing to the VM number that user satisfaction is the cloud computing resources of the user of i, i=1 here, 2 ..., N, and 0 < k ₁< ... < k _n≤ K.Use n _irepresent that in mobile cloud computing service territory, user satisfaction is all numbers of users of i.Satisfaction is that the user of i should distribute k _ithe cloud computing resources of individual VM.

In mobile cloud computing service territory, always have the event of two types:

1) a cloud computing service request of newly arriving, represents with R;

2) user satisfaction is that the cloud computing service of i completes operation, and releases the cloud computing resources shared by it, uses D _irepresent.

Therefore move any event e in cloud computing service territory can use e ∈ { R, D at this ₁, D ₂...., D _nrepresent, all possible states of system represent with S, thus the system state in mobile cloud computing service territory can represent with following formula:

S＝{s|s＝<n ₁，n ₂，...，n _N，e>}.

(2)

2) action collection is set

When an end-user request accesses mobile cloud computing service territory and applies for using cloud computing service (e=R), this moves cloud computing service territory needs to determine whether accept this user request, if accepted, distribute the cloud computing resources of how many VM so should to the user of this request service.For simple meter, with A (s)=0, we represent that this moves cloud computing service territory and refuses a cloud computing service request; Represent that mobile cloud computing service territory have received this cloud computing service request with A (s)=i, and the cloud computing resources distributing to this cloud computing service request is k _iindividual VM, to the satisfaction of terminal user to this cloud computing service can be made to reach i, s represents current system state here.And on the other hand, we represent with A (s)=-1 and terminate to run and state (the event e=D here of cloud computing resources shared by discharging at a cloud computing service _i) under action, namely add up the VM number of existing available cloud computing resources and wait for the generation of next event.Therefore, the action collection of this model is summarized as follows:

$A\left(s\right)=\left\{\begin{array}{cc}-1& e\&Element;\{D_1,D_2,...,D_N\}\\ \{\mathrm{0,1},...,N\},& e=R.\end{array}\right.---\left(3\right)$

3) earnings pattern is set

Based on system state and corresponding action, we can estimate an obtainable income in mobile cloud computing service territory (with z (s, a) represent, a is the action that system is taked each event, comprises refusal cloud computing service request or the cloud computing resources for cloud computing service request dispatching i VM.Under the arrival state of cloud computing service request, a also can be understood as the VM number of the cloud computing resources distributing to each user; But under the state that mobile subscriber leaves after cloud computing service terminates, a=-1 represents the existing available cloud computing resources of statistics.), this income is made up of two parts, and one is the income of system, and another part is the expenditure of system, can represent with following formula,

z(s，a)＝x(s，a)-τ(s，a)y(s，a) (4)

X (s, be a) that system is s in state, when the action of selection is a, the gross income that system obtains, can be expressed as,

$x(s,a)=\left\{\begin{array}{cc}-1,& e=R,a=0\\ U\left(k_i\right),& e=R,a=i\end{array}\right.---\left(5\right)$

Wherein, U (k _i) be Efficiency Function, as shown in formula (1).(s, a) represents at current system conditions to be s, when the action chosen is a, transfers to the service time desired by next system state j τ; Y (s, a) represents to be s at current system conditions, the expenditure when action chosen is a, y (s, a) can measure by total number of the cloud computing resources VM shared by the cloud computing service run, be expressed from the next into:

$y(s,a)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{n}_i{k}_i.---\left(6\right)$

The present invention determines whether determining this request according to the long-term expected revenus that earnings pattern obtains.Long-term gain model of the present invention can obtain the long-term gain of reception, refusal respectively according to earnings pattern, that action a (receiving or refusal) then selecting long-term gain maximum; If have received current request, system state can be upgraded.Can wait for next event, then determine action in that event, and again upgrade system state, the operation gone round and begun again like this is gone down simultaneously.

4) state transition probability is solved

Decision point refers to the time point when any one event occurs, a such as cloud computing service request arrives mobile cloud computing service territory, or one has completed and uses the mobile phone users of cloud computing service to leave this cloud computing service territory and cloud computing resources shared by release.In our system model, due between two decision points time τ (s, a) equal obeys index distribution, therefore, the mean speed γ that event occurs (s, a) can be expressed as,

$\mathrm{\&gamma;}(s,a)=\mathrm{\&tau;}{(s,a)}^{-1}$

$=\left\{\begin{array}{cc}\mathrm{\&lambda;}+\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{n}_i\mathrm{\&mu;},& e=R,a=0\\ & \mathrm{ore}=D_i\\ \mathrm{\&lambda;}+(\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{n}_i+1)\mathrm{\&mu;},& e=R,a=i\end{array}\right.---\left(7\right)$

Thus, time τ (s, expectation discount income z a) (s, a) can be expressed as,

$z(s,a)=x(s,a)-y(s,a)E_s^a\left\{{\&Integral;}_0^{{\mathrm{\&tau;}}_1}{\exp }^{-\mathrm{\&alpha;t}}\mathrm{dt}\right\}$

$=x(s,a)-y(s,a)E_s^a\left\{\frac{1-{\exp }^{-{\mathrm{\&alpha;\&tau;}}_1}}{\mathrm{\&alpha;}}\right\}---\left(8\right)$

$=x(s,a)-\frac{y(s,a)}{\mathrm{\&alpha;}+\mathrm{\&gamma;}(s,a)}$

Wherein, E is under state s, when take action a time, to the average expected time that next event occurs, τ ₁refer to the time experienced to the state that next event occurs from current state.X (s, a) with y (s, a) define at formula (5) and (6) respectively, because the time between two events is continuous print, and we are discrete at the time point doing decision-making, need by changing the income that just can obtain discrete time to the income of continuous time, therefore we represent the discount rate under continuous time with α.We allow p, and (j|s, a) represents that system is at state s, and when the action chosen is a, system transfers to the transition probability of state j.We can derive all state transition probabilities thus.Write to simplify, the symbol that we are defined as follows,

${\hat{n}}_1=<n_1,n_2,..,n_i,..,n_N>$

${\hat{n}}_{2,i}=<n_1,..,n_i-1,..,n_N>$

${\hat{n}}_{3,i}=<n_1,..,n_i+1,..,n_N>$

${\hat{n}}_{4,i,m}=<n_1,..,n_i+1,..,n_m-1,..,n_N>.---\left(9\right)$

When a new cloud computing service request arrives mobile cloud computing service territory, if the at this moment decision-making of system is refusal, so have there is a=0 simultaneously; Or the cloud computing service being i when a user satisfaction terminates to run, this mobile phone users leaves this when moving cloud computing service territory and discharge cloud computing resources, at this moment the number of users accepting cloud computing service inside system reduces and the increase of available cloud computing resources, therefore has in both cases, we can obtain transition probability and are,

$p\left(j\right|s,a)=\left\{\begin{array}{cc}\frac{\mathrm{\&lambda;}}{\mathrm{\&gamma;}(s,a)},& j=<{\hat{n}}_1,R>\\ \frac{n_{i,\mathrm{\&mu;}}}{\mathrm{\&gamma;}(s,a)},& j=<{\hat{n}}_{2,i},D_i>,n_i\&GreaterEqual;1.\end{array}\right.---\left(10\right)$

When a new cloud computing service request arrives mobile cloud computing service territory, if system decision-making at this moment agrees to access and the cloud computing resources being prepared as this cloud computing service request dispatching is k _iindividual VM, so now there is a=i, i=1,2 simultaneously ..., N, in this case, we can obtain transition probability and are,

$p\left(j\right|s,a)=\left\{\begin{array}{cc}\frac{(n_i+1)\mathrm{\&mu;}}{\mathrm{\&gamma;}(s,a)},& j=<{\hat{n}}_1,D_i>\\ \frac{\mathrm{\&lambda;}}{\mathrm{\&gamma;}(s,a)},& j=<{\hat{n}}_{3,i},R>\\ \frac{n_{m,\mathrm{\&mu;}}}{\mathrm{\&gamma;}(s,a)},& j=<{\hat{n}}_{4,i,m},D_m>,n_m\&GreaterEqual;1,m\&NotEqual;i.\end{array}\right.---\left(11\right)$

Fig. 4 gives the mobile system for cloud computing dynamic cloud computational resource proposed based on us and optimizes Decision of Allocation model, state transition diagram during N=2.

5) maximized entire system long-term gain is solved

Thus, according to the definition (SMDP) of semi-morkov decision processes, the maximum long-term discount income that we can obtain the mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation based on SMDP proposed by the invention is,

$v\left(s\right)=\underset{a\&Element;A\left(s\right)}{\max }\{z(s,a)+\mathrm{\&eta;}\underset{j\&Element;S}{\mathrm{\&Sigma;}}p\left(j\right|s,a)v\left(j\right)\}---\left(12\right)$

Wherein (s, a) (j|s, a) respectively in formula (8), obtain in (10) and (11), v (j) represents the overall long-term gain that system obtains when next state j to z with p.

6) optimized decision-making is found

According to the maximum system income that formula (12) obtains, we can find the system decision-making corresponding with this maximum return easily, this decision-making is the optimal decision-making current mobile cloud computing service territory dynamic cloud computational resource being optimized to distribution, and the strategy that the optimal decision-making distributed by the dynamic cloud computational resource optimization of all mobile cloud computing service territories forms is the optimal decision-making strategy of mobile cloud computing service territory dynamic cloud computational resource model of optimizing allocation.

According to step 5) each action is taked to event after obtain corresponding long-term gain respectively, then inside these long-term gains, the action (namely refuse cloud computing service request, or be the VM number of cloud computing service request dispatching cloud computing resources) selecting that maximum long-term gain corresponding.

Claims (3)

1., based on a dynamic cloud computational resource optimizing distribution method of SMDP, the steps include:

1) user satisfaction is divided into N class by cloud computing service domain system, and satisfaction classification is the virtual machine VM number that user's correspondence of i is distributed is k _i; Wherein, 1≤k _i≤ K, K are the VM sum in cloud computing service territory;

2) terminal user sends services request to cloud computing service territory, and application uses cloud computing service;

3) cloud computing service domain system sets up an action collection according to the services request received and current cloud computing service territory state; Wherein, the state s in cloud computing service territory is s=<n ₁, n ₂... n _i... n _n, e>, n _ifor satisfaction classification in cloud computing service territory is the number of users of i, e is the event in cloud computing service territory, e ∈ { R, D ₁, D ₂, D _i., D _n, R is cloud computing service request, D _ithe cloud computing service being i for satisfaction classification completes and releases the VM number shared by it; Described action collection is $A\left(s\right)=\left\{\begin{array}{c}-1,e\&Element;\{D_1,D_2,...,D_N\}\\ \{\mathrm{0,1},...i,...,N\},e=R\end{array}\right.,$ A (s)=-1 represents that cloud computing service terminates to run and discharges the state of shared cloud computing resources, A (s)=0 represents the refusal cloud computing service request of cloud computing service territory, and A (s)=i represents that cloud computing service territory receives cloud computing service request and the cloud computing resources distributing to this cloud computing service request is k _iindividual VM; S represents the current state in cloud computing service territory;

4) for each action in described action collection, the long-term gain in cloud computing service territory is calculated;

5) cloud computing service domain system determines whether to accept current service request according to the long-term gain calculated, if accepted, the VM Resource Allocation Formula choosing the maximum action correspondence of long-term gain is cloud computing service request dispatching VM;

Wherein, formula z (s, a)=x (s is utilized, a)-τ (s, a) y (s, a) calculate for each action a income z (s, a), x (s, a) be state be s, the action of selection is when being a, the gross income that cloud computing service territory obtains, τ (s, a) represent state be s, the action chosen be a time, transfer to the service time desired by next state j; Y (s, a) represent state be s, the action chosen be a time cloud computing service territory expenditure; $x(s,a)=\left\{\begin{array}{cc}-1,& e=R,a=0\\ U\left(k_i\right),& e=R,a=i\end{array}\right.,$ U (k _i) be Efficiency Function; β is the discount rate between two decision points under continuous time, and decision point refers to the time point that any one event e occurs, τ ₁refer to the time experienced to the state that next event occurs from current state; Utilize formula calculate the long-term gain ν (s) in cloud computing service territory during each action a in described action collection, (j|s, a) is state transition probability to p, and j is the NextState in mobile cloud computing service territory, γ (s, a) be the mean speed that event occurs, S is all possible state in cloud computing service territory, the overall long-term gain obtained when v (j) represents NextState j.

2. the method for claim 1, it is characterized in that time τ between two decision points (s, a) obeys index distribution, mean speed γ that event occurs (s, a)=τ (and s, a) ^-1.

3. the method for claim 1, is characterized in that employing one Efficiency Function is to measure the satisfaction of cloud computing user, is divided into N class by user satisfaction.