The Research On Task And Resource Scheduling Of Data Intensive Applications In Cloud Computing

Cloud computing is widely used in many fields currently, especially in scientific computation where many complex tasks of applications are deployed in cloud enviroments. Hence, task scheduling is an important part in cloud sytems. Cloud systems have drawn much attention because it can deliver pay-as-you-go and high available services to cloud users with low costs consumed at the same time. In that case, how to efficiently schedule tasks to virtural machines and allocate computation resources such as network, bandwidth, CPU and so on in cloud systems so as to shorten the completion time of applications is becoming a dominating issue in clouds. Every cloud provider should effectively address this issue to premote service level aggrement(SLA) for cloud users.This paper’s subject is about the data intensive application scheduling and computation resources allocation. The main work that has been done includes the following contents:(1) according to the characteristics of data intensive applications, which is the logical relations between the applications’ subtasks, a deadline based scheduling approach for data-intensive applications in clouds is proposed. This approach consists of two sub algorithms: the partition algorithm for subtask set and the scheduling algorithm for subtasks;(2) based on the unique way of resource allocation in clouds, a three level resource allocation model is proposed. Currently, there are few work about task to virtrual machine scheduling. Therefore, a computation capacity of virtural machines based task scheduling algorithm is also proposed in this paper(CBTSA).(3) there are many data interactions between storage or computation nodes when deploying data intensive tasks in clouds. Hence, a bandwidth based virtual machine placement(BAVP) and migration algorithm is presented in this paper. The proposed CBTSA+BAVP algoroithm is compared with the random placement algorithm and the default algorithm in cloudsim. Related simulations demonstrate its effectiveness.At last, based the above task scheduling and resource allocation algorithms, we have studied one of the most popular cloud platform Open Stack to implement the propsed algorithms above. And they are compared with the random placement algorithm. Experiment results show the superiority of the proposed algorithms.