Computing Resource Scheduling And Allocation For Wireless Small-Scale Cooperative Cloud System

With the fast development of the Mobile Internet,PC is replaced with mobile terminal which is rep-resented by mobile phone.At the same time,the Internet integrates into everyone's life.In the future,the Internet of things which uses 5G as the framework will make people and things linked more closely.While in the real life,mobile terminal is difficult to provide higher computing power and endurance because of its size.Mobile cloud computing is designed to solve this problem.Mobile cloud computing could be divid-ed into two categories,one is the remote cloud,the other is the small-scale cloud.Small-scale cloud can be accessed through a local wireless link which is different from remote cloud.So it has a better flexibility and can be applied to the Internet widely.In the traditional small-scale cloud structure,each access point of small-scale cloud serve only one user.When the user's computational cost is high,it is difficult to improve the performance of the small-scale cloud system.In order to solve this problem,this paper proposes a Wireless Small-scale Cooperative Cloud(WSCC)system.In this system,access points can be connected through the wireless link and organized into a cooperative cloud system to service users.In this process,the user's tasks of this workflow are processed in parallel.It can reduce the user's processing time effectively.This paper mainly discusses the allocation and system queuing process from tasks to access points of WSCC system.Firstly,PCP(Partial Critical Path)algorithm is used to study in the process of task allocation.PCP algorithm is a heuristic algorithm that has two advantages.First,it sets adjacent tasks into critical paths and assigns critical paths into access points,which can effectively reduce transmit data in the system.Second,it can evenly allocate PCPs to access points,which can effectively make the load be more balanced.In the computing resource scheduling process,this paper presents two queuing scheme.One is queuing-before-access-point scheme,the other is queuing-before-system scheme.The WSCC system used former scheme has efficient use of computing resources.In the later scheme,the WSCC system reserves the computing resources in advance for its user,and this lead to a stable performance.Secondly,this paper extends workflow to delay graph which gives model and theoretical analysis of WSCC system.There are two models to forecast the delay.The first is to compute the delay under the assumption of computing resource shortage.The second is to compute the delay under the assumption of communication resource shortage.The congested access points or wireless links of the two assumption can be predicted by calculating the longest path of the delay graph.This paper also designs two equations.The WSCC system's performance of the two models in different arrival density can be obtained by solving the two equations.Then,the average dwell time of the users in the theoretical analysis is obtained by summing up the two models' predicted values with weight.Finally,the simulation platform is designed to test the performance of WSCC system.Simulation plat-form is divided into five modules:the system's input module,task allocation module,user accessing module,computing resource processing module and communication resource processing module.With the used of the simulation platform,we can compare the performance of the two queuing schemes,compare the simulation results with the theoretical analysis,compare the performance between non-cooperative system,WSCC sys-tem based on PSO algorithm and WSCC system based on PCP algorithm,test the influence of each variable by using the control variable method.We can conclude the simulation results that the queuing-before-access-point scheme gives a better performance than the queuing-before-system scheme,the small-scale cooperative cloud system can achieve a better performance than the small-scale non-cooperative cloud system,the PCP algorithm give a well performance with less effort,the theoretical analysis based on delay graph can predict the system performance within accepting error.