An Adaptable Scheduling for Self-Reconfigurable Objects

The execution environment of a real-world distributed system can change dramatically in terms of CPU workload, memory utilization, network traffic, etc. Therefore, an object-oriented distributed environment called Juice was developed in our previous research. Applications developed with Juice have the capability of reconfiguring itself during runtime in order to adapt changes of the execution environment. In this work, we propose a base level adaptable scheduling for the execution of self-reconfigurable objects in Juice. The adaptable scheduler can dynamically change the scheduling policy to meet the user's performance target when the execution environment changed. Evaluation results show that in a low CPU workload environment, FCFS (First Come First Serve) policy has a similar performance with SJF (Shortest Job First) policy and LJF (Longest Job First) policy with lower scheduling overhead. However, when the CPU workload becomes heavy, SJF or LJF has a better response time depending on the ratio of long jobs and the number of worker threads of Juice.