SHARP: a new real-time scheduling algorithm to improve security of parallel applications on heterogeneous clusters

This paper addresses the problem of improving quality of security for real-time parallel applications on heterogeneous clusters. We propose a new security- and heterogeneity-driven scheduling algorithm (SHARP for short), which strives to maximize the probability that parallel applications are executed in time without any risk of being attacked. Because of high security overhead in existing clusters, an important step in scheduling is to guarantee jobs' security requirements while minimizing overall execution times. The SHARP algorithm accounts for security constraints in addition to different processing capabilities of each node in a cluster. We introduce two novel performance metrics, degree of security deficiency and risk-free probability, to quantitatively measure quality of security provided by a heterogeneous cluster. Both security and performance of SHARP are compared with two well-known scheduling algorithms. Extensive experimental studies using real-world traces confirm that the proposed SHARP algorithm significantly improves security and performance of parallel applications on heterogeneous clusters