Cloud Scheduling System Considering Reliability and Co-Resident Attack

Abstract

For developing an efficient job assignment and resource management mechanism considering reliability and cost, this paper proposed a biologically-inspired technology - Bionic Cloud Scheduling System (BCSS). The BCSS is designed to consist of a global job assignment agent, multiple local resource management agents, and large-scale infrastructure. The primary innovation of the BCSS is that it takes the global agent, local agents, and infrastructure as the brain, autonomous nerves, and cyber neurons, respectively. Meanwhile, this paper also presents a realization of the BCSS. Cyber neurons are realized as theoretical models evaluating reliability and service performance while capturing co-resident attack factors in real-time. The autonomous nerve processes an autonomic resource management mechanism to achieve dynamic balances between reliability and cost. The brain is capable of job assignment to ensure global optimization on an object of minimizing penalty and cost. Experimental results illustrated the important correlation between reliability and co-resident attack factors. It also demonstrated that the BCSS could optimize a perception metric related to reliability, co-resident attacks and cost factors.