Multi-Installment Scheduling for Networked Computing Systems with Server Release and Offline Times

Effective task scheduling has become one of the most crucial issues for achieving good performance on networked computing systems, especially for nowadays big data related applications. Most existing scheduling models assume that plenty of servers are available when users submitting their workloads to the computing system and all servers involved in workload computation are able to stay online forever. In reality, servers, however, may have arbitrary unavailable time periods. Hence, if we inadvertently assign tasks to servers without considering the availability constraints, inclusive of server release times and offline times, some servers would not be able to start workload computation immediately and could not finish their assigned workloads on time. Thus all the unfinished workloads need to be reassigned to other available servers resulting in an inefficient time schedule. In this paper, we propose a novel periodic multi-installment task scheduling model and design a time-efficient genetic algorithm to derive an optimal load distribution strategy. Our experimental results show that the proposed algorithm adapts to minimize the processing time by at least 35% compared to other strategies.