DYNAMIC SCHEDULING WITH LOAD BALANCING IN HARD REAL-TIME DISTRIBUTED SYSTEM

Abstract

Many tasks, such as those found in nuclear power and process control applications, are inherently distributed and have severe real-time constraints. Because, their execution deadlines must be met, these tasks are said to have hard real-time constraints, A scheduling algorithm that schedules arriving tasks dynamically and efficiently on the nodes of a loosely coupled distributed system, to meet real-time requirements has been presented in this dissertation. In some cases it is possible that when the task arrival rate at one node is high, it may be low for another node. The busy node must offload some of its tasks onto the idle nodes, thereby completing execution of these real-time tasks before their deadline. To do so the heavily loaded node must find .a suitable host for remote execution. A scheme for finding best remote host quickly, is also presented (here centralized architecture is assumed). The performance analysis is done by simulating the behavior of algorithms under different system load conditions such as homogeneous and non-homogeneous arrival rates. The results shows the suitability of the algorithms to real-time applications.