Scheduling Strategy Study And Performance Analysis In Cloud Computing

As a focus problem,resource scheduling not only directly improves the performance of cloud computing,but also satisfies the SLA-based QoS described by consumers.The properties of cloud computing,such as large-scale,dynamics,heterogeneity,and diversity,present a range of challenges for the algorithm research and performance evaluation on resource scheduling.This paper provides improved and effective strategies,and use evaluating models to solve resource scheduling problems in cloud computing.The work is summarized as follows:Firstly,considering resource heterogeneity,bandwidth variability and QoS requirement diversity in cloud computing,the task scheduling problem is formulated,and a PSO-based adaptive multi-objective task scheduling(AMTS)strategy is proposed.AMTS is advanced to achieve the optimal resource utilization,task completion time,average cost and average energy consumption.In order to maintain the particle diversity,the adaptive acceleration coefficient is adopted.Experimental results show that AMTS can obtain quasi-optimal solutions than that of genetic algorithm(GA)for task scheduling in cloud computing.Secondly,Petri net theory is first used in modeling and performance analysis of cloud computing.The study applies markov modulated deterministic process(MMDP)on describing the task scheduling in cloud environment,meanwhile,deterministic and stochastic Petri net(DSPN)is applied to model and analyze fair scheduling based on max-min fair scheduling(MMFS)algorithm and priority-based MMFS(PMMFS).MMFS and PMMFS can be expressed by setting enabling predicates and random switches in the DSPN model.In addition,decomposition and iteration techniques are introduced to analyze multiple user's scenario,which can reduce state space and avoid state explosion.The study emphasized on average performance of fair scheduling,such as average throughput,average queue length and average delay.Thirdly,focus on the deficiencies of modeling and evaluating task scheduling algorithms in cloud computing,propose a novel dynamic scalable stochastic Petri net(DSSPN).DSSPN is based on SPN and Stochastic Reward Net(SRN),and offers the theoretical foundation for subsequent study.Furthermore,we elaborate the dynamic property of DSSPN,and demonstrate some properties.In addition,we present the classified fair scheduling(CFS)with considering jobs diversity and resources heterogeneity,and then evaluate the performance of the fair scheduling and CFS based on DSSPN.Finally,with the increasing call for green cloud,energy consumption consumed by the cloud datacenters has gained widely concern for cloud service providers.It can not only reduce the operating cost,but also improve datacenter sustainability.Considering energy consumption,resource diversity and virtual machine migration,EAVMS algorithm combined with EAVMM algorithm are proposed and applied to guide the task scheduling and in energy-aware migration-enabled clouds.In addition,DSSPN is also applied to model and evaluate this task scheduling flow,which is also further verify the availability and feasibility of DSSPN.This paper studies the dynamic optimization scheduling in cloud computing,and put forward some scheduling algorithms such as AMTS,CFS,EAVMS and EAVMM in order to achieve better performance in some aspects.In addition,the main contribution of this paper is the application and development of Petri net theory in cloud environment,which can be improved continuously in practical application.