A distributed look-ahead algorithm for scheduling interdependent tasks

Autonomous Decentralized Systems concurrently work on different types of jobs which consist of interdependent tasks ("steps"). Steps are characterized by their service requirements. It is the load balancing problem to increase the system throughput by reducing contention between steps accessing the same resources. In this paper, we propose a dynamic decentralized look-ahead scheduling algorithm and a cooperation protocol. The goal is to utilize information about the internal job structure concerning future service requirements and system state information for dynamically arranging schedules such that jobs can take advantage of inevitable waiting times of others. We evaluated the proposed algorithm by simulation. Waiting time reductions in sample configurations of up to 75% for single job types at the expense of only slightly worse response times for other job types compared to systems not applying the algorithm proof the algorithm’s success.