Research On Tasks Scheduling In Cloud Environment

Ultra-large-scale, dynamic and heterogeneous are the most important characteristics ofCloud Computing Systems, and tasks in Cloud environment are triggered at any time, so thesystem needs to manage and schedule tasks frequently. How to make efficient scheduling ofheterogeneous nodes, how to ensure each task is completed quickly, while cloud computingsystem is maintained at a relatively load balancing state, which is a hot topic of cloudcomputing. There are certain aspects of the same characteristics in task scheduling amongcloud computing, traditional distributed computing and grid computing. As cloud computingand traditional distributed computing and grid computing has certain aspects of the samecharacteristics in task scheduling. Task scheduling algorithm of cloud computing system islargely inherited distributed computing systems and grid computing systems. As a businessmodel, task scheduling strategy of Cloud environment is not only taking account into the taskexecution efficiency and also business benefits of cloud service providers which include cloudcomputing system resource utilization and energy consumption. So far, there are someresearches in task scheduling of Cloud environment. Therefore, the study of task schedulingfor cloud environment is the key issue for improving service capabilities of cloud computingsystem, which has a very important practical significance and theoretical value.This paper focuses on the task scheduling strategy of cloud environment, and made acomprehensive and systematic analysis and research on it. This paper mainly includes fouraspects, which are execution efficiency, cloud computing system load balancing problem,multi-dimensional resource utilization for virtual machine deployment issues and greencomputing issues. The main contribution of this paper and research results are as follows:1) For the heterogeneity of physical hosts in Cloud Computing system and thecomplexity of user tasks, it is really a big issue that how to set tasks mapped onto physicalhosts of system ensuring tasks can be completed as soon as possible and balancing the load ofsystem. We proposed a heuristic task scheduling strategy of cloud environments based on antcolony algorithm. We construct a task scheduling model based on task performancerequirements and resource needs, turn task deployment problem into a group of ants movingon a bipartite graph, and we construct the corresponding scheduling assessment strategies.The optimum task allocation strategy will be found progressively by analyzing the ants walking path. Compared with traditional methods, the proposed method may meet differentuser quest QoS, and maximize the efficiency of cloud computing systems and try to ensurethat load balance of system.2) In cloud environment, any network node can trigger tasks which may be unevenlydistributed. If we only consider the single-task performance, it will lead to uneven loaddistribution. How to compromise load balancing and system performance is a hot topic in taskscheduling strategy of Cloud environment. We model task allocation process and proposed a2-step load balancing algorithm of Cloud environment which include task assignment andVM deployment. In step-1, we assign tasks on VMs which can meet the requirements of eachtask. In step-2, a new VM will be created if no VMs can meet task’s requirements, and virtualmachine deployment strategy is designed based on the goal of load balancing. Experimentalresults show that the algorithm will ensure load balance and reduce the average waiting timesignificantly.3) In cloud environment, heterogeneity will lead to greater resource utilizationdifferences between different computer nodes, even which lead to unbalance of whole Cloudcomputing system. To improve the utilization of resources in cloud computing system, wepropose a virtual machine deployment strategy which aimed for multi-dimensional resourceand heterogeneous system based on Group Genetic Algorithm. A hierarchical characteristicchromosome encoding rules is constructed based on characteristics of computing resourcessystem. Gene box represents physical host, and gene represents virtual machine. Optimizationtarget is minimization of number of open physical hosts and maximization physical hostresource utilization. Experimental results show that the algorithm will ensure the utilizationrate and load balancing of physical host, and minimize the number of open physical hosts.4) With the size of the Coud continues to expand, energy consumption is graduallyincreasing for Cloud Computing System, which become a big burden of carriers. We proposedan Energy-Aware Virtual Machine Migration strategy using Particle Swarm Theory aimed atpower consumption of Cloud Computing data center. We modeled user request, virtualmachine and physical host energy consumption. The method improved systemmulti-dimensional resource utilization by migrating virtual machines and reduced powerconsumption by close idle physical hosts. Experiments show that resource utilization isimproved significantly and energy consumption decreased obviously after the algorithm isexecuted.Through these studies, this paper optimized the existing task scheduling algorithm fromfour aspects which are task performance, resource utilization, system load balancing and energy consumption. These work for the end user can optimize their user experience, for thecloud service providers can reduce their development and maintenance costs and for cloudsystems operators can reduce their operating costs and management costs. In summary, thisresearch can further promote the development of Cloud computing, and produce considerableeconomic benefits.