Research Of Cloud Task Scheduling Based On Antcolony Algorithm And DAG Workflow

As a new business model, cloud computing promote the development ofindustrial production. Cloud computing provide a variety of services for users in atransparent manner. Users will be able to obtain a wide range of services according totheir needs with no need to understand the technical realization of cloud computingplatform. In cloud environment, the allocation of virtual resources to all types of usersis a key technical problem. A good resource allocation strategy should meet userdemands with improving the system operating efficiency. Therefore, cloud taskscheduling research has practical significance.In this paper, we study the key technologies of cloud computing in-depth,especially the cloud computing task-scheduling algorithm. As the deficiencies of theexisting scheduling algorithms, we propose an ant colony optimization-schedulingmodel to solve the scheduling problem with different task model. Firstly, we use antcolony to build solution of single-task system, then use ants search solution of DAG(directed acyclic graph) task. The major work of the paper is as follows:(1) In single-task scheduling system, tasks are independent with each other;assign tasks to virtual machines can be seen as a multi-objective optimization problem.Ant colony can solve this problems with continuous iterative based on a positivefeedback mechanism between the tasks. The paper presents a method for solving tasksscheduling by building intelligent "artificial ant", making ants quickly converge to anoptimal solution. In this paper, we use CloudSim, a high-performance cloudcomputing simulation platform, to do simulation experiments and compare ant colonyalgorithm with FCFS (first come, first served) and greedy algorithm.(2) In the actual situation, there have associations between tasks. The paper useDAG workflow model to describe this complex task scheduling system. In order tosolve such tasks scheduling problem, this paper firstly proposes a priority-basedscheduling algorithm by giving priority to the tasks and dynamically allocating virtualmachines. On the basis of this algorithm, we propose an optimal ants systemintegrated with priority scheduling algorithm, which combines the advantages of antcolony algorithm and priority-scheduling algorithm.