Optimizing software rejuvenation policy based on CDM for cloud system

摘要

The ever-increasing cloud-based systems are suffering progressive performance degradation due to internal accumulated errors and external security attacks. Software rejuvenation is a preventive maintenance policy to solve this problem. However, an ineffective rejuvenation policy usually causes high resource consumption and system cost loss. To guarantee the high availability and minimize the maintenance cost in the cloud-based system, a cost-effective rejuvenation policy is proposed based on cumulative damage model (CDM) in this paper. With our proposal, prior to system failure diagnosis, the system is managed to be rejuvenated either at a predetermined time or at a damage threshold level, whichever occurs first. Further, the system cost rate function is deduced, and an optimal rejuvenation period is derived aiming to minimize the maintenance cost. Experimental results show that the proposed rejuvenation policy is highly effective in term of great reduction on system cost rate.