摘要

In cloud storage, how to effectively schedule user's request to specific disk becomes to be a hot topic. Considering the new requirements related to energy-aware scheduling algorithms proposed by cloud storage, e. g. sensitivity to request's response time and limited space left for dynamic optimization, this paper proposes a layout-based scheduling approach for virtual disks. The disk array is dynamically partitioned into two sets, working set and standby set respectively using this approach, and requests are mainly addressed by working set. Additionally, according to real-time workloads, the layout of virtual disks is optimized with standalone virtual disk as a unit. Results of experiments show that this approach can cut down the power of disk array, effectively alleviate the problem of prolonged response time, and achieve higher level of load balancing.

全文