On the analyses of scheduling factors for distributed client/server environment

Abstract

Scheduling for resource sharing and load balancing between distributed resources such as server processes has been one of the main issues in distributed client/server environment. A scheduler with a scheduling algorithm as a broker plays an important role of mapping clients into servers. An optimal scheduling algorithm can be obtained only if we define all parameters related with scheduling and the complete knowledge of the behaviour of a task is available, which is very difficult. We can improve scheduling algorithms if there is much useful information. By estimating maximum service time allowed by a scheduler without decreasing system performance, the scheduler can be more effectively constructed. When we plan to provide services in distributed client/server environment, there are many considerations in configuring the system. One of them is to decide the number of servers. It is more important in static scheduling algorithm. If we can decide optimal number of servers, we can establish effective distributed system with high system performance and server utilization. In this paper, we define two primary factors: 1) maximum service time allowed by a scheduler and 2) optimal number of servers. They can improve a scheduling algorithms and affect effective resource sharing and server utilization, which results in fair load balancing, high system performance, and throughput in distributed system. We design the distributed client/server processing environment and analyze these factors based on queueing theory.