Job Scheduling in Homogeneous Distributed Systems

Today’s world demands a lot of computing power for many different applications. Distributed systems offer this with their advantages. High-performance computing clusters are suitable for running different kinds of jobs like tightly coupled parallel and distributed applications. The queuing system is used to organize tasks and allocate adequate resources at appropriate time intervals. One of the fundamental elements in scheduling tasks is to determine the type and characteristics of tasks that will run in a distributed system. The scheduling algorithm is responsible for the proper assignment of these tasks, to the available resources of a particular node. The most important advantage of using the job scheduler in a homogeneous environment is the fact that the scheduler can omit checking of various parameters. Job scheduling aspects in homogeneous HPC clusters environments is presented in this paper. The grid engine was used as a case study for testing common used algorithms for job scheduling. This example showed the problems that may occur when scheduling tasks, depending on the type and quantity of tasks running. The basic algorithm used in this case does not generally meet their function. Complicated cases require more complex algorithms, taking into consideration proper resources utilization