Standard deviation of probability of processor availability based task scheduling algorithm

Abstract

Cloud [2] is an interconnection of clusters of different heterogeneous processors that provides different services such as SaaS, PaaS, IaaS etc., to different users. User's jobs are diverse and have different requirements. Jobs can be represented as Directed Acyclic Graphs and scheduling these jobs is a known NP-Complete problem. There are many heuristics developed earlier for job scheduling, that work differently in different environments with different requirements. A new heuristic SDPA, using Standard Deviation of probability of resource availability is developed for scheduling tasks which reduces makespan. It works in two stages. During first stage jobs are arranged based on the upward ranks and during the second stage jobs are allotted to the processor based on the standard deviation of probability of processor availability. Performance metrics such as Speedup, SLR and makespan are used to compare the performance of the new heuristic with existing one.