ABSTRACT

Real-time computations require exact bounded response times. For relatively simple models of computation, it is possible to determine conditions under which it is theoretically possible to guarantee that an invocation of a task will complete its execution. But it is very much complicated to guarantee a response time for the periodic tasks that execute in a distributed system application. In a distributed system, the applications are aligned as a set of dependent tasks. Here, the time required for a message to pass from one task to the other is a function of the individual timelines. Each task further constitutes of subtasks related by data dependencies. Also, tight bounds are essential for the tasks because of the dynamic nature of the distributed systems. As the distributed systems have the hard real time deadlines, the task executions have to be completed before the specified deadline. Hence the estimation of the worst case latency for a task link is essential. And, as the task structure and the application system model are well known in advance, for the synthesis and verification of applicationspecific distributed systems this estimation is quite important. This paper proposes an approach to estimate the worst case latency of periodic tasks in a real time distributed system.