Task Scheduling Strategy Based On Tree Network In Cloud Computing Environment

With the popularity of the Internet，the modern society face the blast ofinformation. The main features of data nowadays are diverse, heterogeneous and large.How to manage large volumes, complicated and mutable data and how to find theinformation we interested in large number of data, which has became a difficultproblem of data process. According to System Architecture, cloud computing can bedivided into: the layer of Access; the layer of application interface and the layer ofinfrastructure management. The role of layer of infrastructure management is tomanage the resource, use appropriate algorithm to process data and schedule tasks,provide efficient services, it play a vital role of the whole system.Cloud computing is a process model based on network, it provide the software,hardware and information to user. Its core idea is to entire network as a virtual supercomputer, It firstly cut a large computing task which is difficult for a single PC intoseveral sub-tasks, secondly transfer them to a number of servers through the network,lastly the results data were aggregated and returned to the user.The task scheduling strategy has a great impact on system performance in cloudcomputing environment. In traditional distributed computation such as grid computing,scheduling algorithms for tree network were always based on traditional six modelsfor server，especially full overlap, single port. These six models have one commonpoint that all the servers can’t support parallel multiple tasks. However, these sixmodels are all not appropriate for cloud computing environment nowadays, for themulti-core technology and virtualization technology has been more maturity, oneserver now can parallel process multiple tasks. In recent years, some scholars havedone a lot of researches of task scheduling in cloud computing environment. However,these researches, in terms of network topologies, were almost based on opaque cloud.When the participants of the cluster were organized into a tree structure, appropriatetask scheduling strategy tailored to the situation of multitasking is still lacked.This paper focused on the feature of cloud computing environment, proposed anew framework of task scheduling strategy for tree network—Independent Task onTree Network in Cloud Computing. This framework is still based on master-slavemode, but different from the past, it breaks the limitation of the six models. Eachnode’s system platform of this architecture are based on parallel processing of multi-tasking, so the scheduling framework is fit for the mainstream operating systemnowadays. The scheduling framework starting from single-node performance changesas well as changes in the net work transmission line bandwidth, select the appropriatenode to allocate task dynamically. This paper, firstly, consider about how to use thescheduling framework in two level trees and give the specific implementation process.Then use this as the basis, extending this framework to depth greater than two treenetwork. In the discussion of multi-layer tree structure, we designed two differentalgorithms based on different ideas. One is grant the master node all the right tochoose target node for each task; and in the other one, tasks are allocated by using amanner of decentralization of authority between master and slaves, it give the right oftarget node selection to the slave. The experimental portion of the article focused oncomparing the pros and cons of the two algorithms of multi-layer tree structure.Firstly, we compare the two algorithms from the rationality of target node selection. Inthis part, we build the experimental environment by using simgrid which is adistributed computing simulator working in Linux. Through the experimental data wecan get which one is closer to the optimal solution. Secondly, we discussed about thenetwork overhead brought by the two algorithms themselves. We get the conclusionby mathematical formula. Lastly, we compare the Time consumption of calculate forthis two algorithms. This scheduling framework provided an available solution of theproblem of task scheduling in tree network.