Smart Scheduling Strategy for Lightweight Virtualized Resources Towards Green Computing

Abstract

Modern cloud orchestrators are generally designed to make efficient use of resources in the data center, by consolidating the servers workload. Recently, energy efficiency has become critical factor to sustain the growth of cloud services; thus, more effective resource allocation and management strategies are required. The situation is exacerbated by introduction of HPC-oriented cloud services, where other aspects of the application execution are critical, such as the minimisation of the makespan. Although a short makespan allows for a rapid application execution, often the overall energy consumption of the whole cluster suffers, growing out of all proportion. Starting from the growing attention paid in recent years to the concept of “green computing” (or ICT sustainability), in this paper we propose a different type of resource scheduler, whose main objective is to maximise the (energy) power efficiency of the computational resources involved, while taking into account the overall application execution time. An artificial intelligence (AI) technique, in the form of population-based evolutionary algorithm, was used to develop the proposed scheduler, in order to find the best possible combination between tasks to be performed and usable nodes able to guarantee lower (energy) power consumption and, at the same time, the fulfilment of possible constraints related to tasks’ execution. This paper focused on the implementation and evaluation of an evolutionary algorithm for efficient task scheduling. Experimental evaluation of such algorithm is discussed.