An important part of a distributed system design is the workload sharing among the processors. This includes partitioningthe arriving jobs into tasks that can be executed in parallel, assigning the tasks to processors and scheduling the task execution on each processor. In many system contexts jobs must depart in the order of their arrival, hence the resequence problem is involved. Inthispaper we examine the efficiency oftwotask routing strategies — one static and one adaptive — and three nonpreemptive task scheduling policies in conjunction with job resequencing before departure. It is shown thatthe adaptive task routing strategy outperforms the static one and that when adaptive task routing is applied, the scheduling strategy affects marginally system performance. The minimum resequence delay is achieved with probabilistic task routing and FCFS task scheduling.
Anthony Karageorgos, Helen D. Karatza