Processing jobs using speed-scaling function to minimize flow-time plus energy in data centers

This paper evaluates the performance of online scheduling for the objective of minimizing the flow time plus energy usage under dynamic speed scaling model and using arbitrary power function in data centers. An algorithm (Jyoti Algorithm) JAlg(sRPT, P∼∗((3Aj/ 2) + (1/ 8)
^2A
)) is proposed which uses Shortest Remaining Processing Time (SRPT) for scheduling jobs and processes jobs at speed such that the consumed power depends on the queue length (i.e. number of active jobs), rather than the remaining work of active jobs. It is applicable in both the settings: clairvoyant and nonclairvoyant. The amortized potential function analysis is used to calculate the competitive ratio of JAlg. The competitive ratio of JAlg is (41/16) against an optimal offline adversary.