Moldable Load Scheduling Using Demand Adjustable Policies

Abstract

Workload distribution among processors is one sided task. Whereas consistent management of processor availability to bulk job arrival is an aspect of resource management. Parallel systems where high probability of infinite job arrivals with varying processor demand requires a lot of adjustment efforts to map processors space over job space .Each job has different required characteristics like no. of processors etc. But the number of available resources is of different characteristics. Particular characteristic processor demanded by a job usually is not available. Such case scenarios are then adjusted to adapt moldable parallel characteristics. Rigid based approaches considered as static demand fit allocation schemes where the job is considered to be active task only when scheduler satisfied the processor demand. Current research focuses on demand adjustment schemes by considering synthetically generated work load and processor availability map with discrete clock frequency. Illustrations produced on the basis of simulation study about demand adjustment schemes consisting static and dynamic approaches with the aim of consistent processor availability fit i.e. processor offered space. Idea behind such experimental study is to analyze various scheduling algorithms along with different performance parameters managing best processor space.