An “on/off” model for energy-efficient scheduling of workflow applications in computational grids

A computational grid is a high performance computing system consisting of computer resources distributed over multiple locations and connected via computer network. One of many possible types of applications executed in computational grids is known as workflow applications. These applications consist of multiple computational tasks, which are precedence related, and usually process huge data files. Both the resource allocation and task scheduling according to a given criterion are important steps, which may result in better execution of the workflow application. In recent years one of the mostly used objectives is the minimization of the energy consumption. Sometimes a good policy for such problems is to execute all the tasks as quickly as possible and then turn off the resources used. A so-called “on/off” model is proposed for this problem. Some resource allocation algorithms as well as scheduling ones are developed for the proposed model of the problem in order to find a set of feasible allocations of resources to tasks and the best schedule taking into account consumption of the energy.