MQDS: an Energy Saving Scheduling Strategy with Diverse QoS Constraints Towards Reconfigurable Cloud Storage Systems

Xindong You,Dawei Sun,Xueqiang Lv,Shang Gao,Rajkumar Buyya

DOI: https://doi.org/10.1016/j.future.2021.11.025

IF: 7.307

2021-01-01

Future Generation Computer Systems

Abstract:Nowadays, cloud storage systems are usually constructed by heterogeneous disks with reconfigurable speed for energy efficiency. Scheduling tasks with diverse QoS (Quality of Service) requirements to disks with different speeds is seldom considered in state-of-the-art mainstream disk scheduling algorithms. This paper proposes a multi-QoS disk scheduling strategy (MQDS), aiming to support energy-saving and diverse QoS constraints towards reconfigurable heterogeneous cloud storage systems. Three algorithms are designed in MQDS: Time Prior Disks Algorithm (TPDS) for time-sensitive tasks, Cost Prior Disks Algorithm (CPDS) for cost-sensitive tasks, and Benefit Function-based Disks Algorithm (BFDS) for tasks with time-varying QoS requirements. CloudSimDisk simulator is extended to evaluate the response time, energy consumption and cost expenditure performance of the proposed algorithms, for comparison with the extended Robin Round Disk scheduling algorithm (E-RRDS). Extensive experimental results demonstrate that the three proposed algorithms outperform E-RRDS. TPDS consumes shortest response time and CPDS has the least cost; and BFDS achieves balance between TPDS and CPDS, which provide more options to users. It is worth mentioning that all the algorithms in MQDS are more energy efficient than E-RRDS. As a whole, the proposed MQDS is energy efficient and able to accommodate diverse QoS constraints well.