Research On The Taskflow Multi-objective Scheduling Optimizing Technology In Cloud

Cloud computing has become a more and more popular computing paradigm in the field of high performance distributed computing.Because it provides the service of on-demand access to shared resources in a manner of self-service,dynamic scalable and quantifiable way.At present,the cloud computing is still in an initial stage,which needs a lot of research on many subjects to get its benefit.Among them,one of the most important issues is how to make efficient scheduling optimization which meet multi objective resources criterion.The scheduling problem in the cloud environment belongs to the NP-hard issue,which need to spend a lot of time to find an optimal solution.It can be proved that there is no algorithm to solve these problems in polynomial time.In the cloud environment,it is possible to find the optimal solution or near optimal solution in the shortest possible time.The technologies based on meta-heuristic algorithm have been proved to be able to obtain the approximate optimal solution for the above problem in a reasonable time.Generally speaking,there are different factors and "participants" in the heterogeneous cloud environment.Each of them observes and solves the problem from different angles.Scientists are usually devoted to minimize the application execution time and reduce the economic costs,while the administrator is always concerned about how to maximize the use of resources,reduce energy consumption as much as possible,improve job throughput and ensure that the system user fairness and so on.Therefore,multi objective scenarios often occur in the scheduling problems,especially sometimes many of the important goals are conflicting.For example,a powerful processor usually takes a higher price from the cloud service provider but would consume more energy.In addition,it may become unreliable due to a large number of requests and user contention.Nowadays,the related task scheduling in cloud environment is mostly confined to only one optimization goal(usually the completion time),while only few of them try to balance the optimization between two criteria.A general scheduling framework(high efficiency and low time complexity)and heuristic algorithms to optimize and dynamically balance multiple objectives are still missing.In this paper,we study the task flow multi objective optimization scheduling technology in the cloud environment.In the background of energy saving and economic cold wave in recent years,the multi objective optimization technology in the cloud environment has become the key content in the related area.In particular,introducing the economic parameters as one of the multi objective criteria is an emerging research direction.The purpose of this paper is to study how to dynamically process the IV massive resources in the cloud in a manner of saving energy consumption while reducing the cost of management operation,save more economic cost,maximum benefit with minimum investment and to achieve a win-win situation between users and cloud providers,which have important worth of research and application.For some of the above problems,this paper carried out a preliminary study and exploration,the contributions are as follows:Firstly,we propose the IaaS cloud energy efficiency model and energy efficient solutions.In the cloud platform level,we designed two time zones based energy management strategy.We verify it in the simulation environment and then deploy them on the real IaaS cloud platform.In this part,different task types(CPU intensive and I/O intensive)are designed to verify the applicability of this strategy.Through experiments,it is proved that this strategy can make the data center show better energy saving characteristics with the average savings about 12.98% of the total energy consumption.Secondly,we aim to develop an adaptive task scheduling strategy.In particular,we first model the virtual machine energy from the perspective of the cloud task scheduling,then we propose a genetic algorithm to achieve adaptive regulations for different requirements of energy and performance in cloud tasks(E-PAGA).From the extensive experiments,we pinpoint several important observations which are useful in configuring real cloud data centers: 1)we prove that guaranteeing the minimum total task time usually leads to low energy consumption to some extent;2)we must pay the price of the sacrificed performance if only taking into account the energy optimization;3)we come to the conclusion that there is always an optimal condition of energy-efficiency ratio in the cloud data center,and more importantly the specific conditions of the optimal energy-efficiency ratio can be obtained.Thirdly,aiming at the key problems of efficient scheduling of large number of task flow applications in the cloud environment,we propose a spark based optimization framework,construct the time and economic models based on the two operators in Spark(transformation and action),and then study the scheduling optimization of economy performance tradeoff by using the elastic resources in cloud.In the comprehensive experiments,we design some greed,probability and other search algorithms and several interesting common features are summarized.The practical significance of our research findings lies in that it can be used by the cloud service providers to provide the opportunity to choose the appropriate economic-time schemes to the cloud consumers.Finally,we focus on optimizing the scheduling process in dataflow and dividing the optimization objectives into user metrics(the makespan and economic cost)and indicators of cloud systems(the network bandwidth,storage constraints and system fairness)which is inspired by the reality.An efficient multi-objective game algorithm(MOG)has been proposed,which is formulated as a new cooperative game.The MOG method is able to optimize the user metrics while satisfy the constraint of the system metrics and ensure the efficiency and fairness of the cloud resources.Through comprehensive experiment,it has been proven that compared with other related algorithms,the proposed MOG method has obvious advantages in terms of the algorithm complexity(46)(l?K?M)(improvement of magnitude),the results quality(optimum in some cases)and the system level fairness.