Task Scheduling in Cloud Environments

Abstract

In distributed heterogeneous computing environments, such as cloud computing, hundreds of online users may individually or collectively submit thousands of jobs anytime and anywhere to dynamic resources. Given the large number of arriving tasks resulting from dividing these bulk-submitted jobs, and considering the amount of data being used in these tasks, optimal scheduling can become a serious problem in cloud environments – where stochastic jobs contend for limited computer and storage resources. Task scheduling optimization plays a crucial role in minimizing service response times and costs while also improving quality of service (QoS) but, in a dynamic cloud environment, this is a non-linear multi-objective NP-hard problem. The main concerns with dynamic scheduling are how to adapt to the inherent uncertainty and how to cope with conflicting objectives, such as minimizing task transfer times, task execution costs, power consumption, task queue lengths, and response times. Evolutionary algorithms can solve such problems by significantly reducing the complexity of the search space to ensure acceptable runtimes for the scheduling algorithm. This chapter reviews the recent literature on models that use evolutionary algorithms, such as particle swarm optimization (PSO) and genetic algorithms (GA), to optimize task scheduling in cloud environments. The analysis is based on the type and number of objectives, as well as the efficiency and reliability of the developed models and applied algorithms.