A QoS-constrained scheduling for access requests in cloud storage

The booming industry of cloud computing encourages more and more people to put various types of data into the cloud. It results in a more frequent access to the data that stored in the cloud. Scheduling these requests to ensure the performance of cloud storage and the satisfaction of users becomes the key issue in cloud computing. Unfortunately, existing work mainly focus on task scheduling and resource allocation, which cannot be applied to scheduling access requests for data directly. Moreover, some of the current works only consider single metric such as the shortest path or the minimum cost. In this paper, we propose a QoS-constrained scheduling mechanism for access requests in cloud storage. We firstly design an algorithm with static load. This algorithm considers both performance QoS and trust QoS for requests to improve resource utilization and ensure the security of both requests and virtual machines in open storage architecture. Experiment results show that this QoS-constrained scheduling mechanism is capable of achieving superior performance and good scalability than all baseline approaches. Meanwhile we notice that with static load global utility will be reduced when several requests in the queue compete for a limited number of resources. To avoid being over loaded when scheduler puts too many requests on some virtual machines, we allow migration in system and apply a dynamic load balance strategy by minimizing global utility loss in the scheduler. The results of our experiments show that dynamic load balance strategy makes global utility increased. It also ensures fairness between all requests and achieves an approximate optimal result.