Microscheduling and Randomization Strategies in Distributed and Cloud Computing
Victor V. Toporkov, Dmitry Yemelyanov · 2020
In this work, we study heuristic and randomized approaches for parallel jobs execution in high performance computing environments with heterogeneous resources. Traditional process of a job-flow scheduling implies many requirements for jobs execution criteria and priority and, thus, usually cannot be changed. However, special microscheduling heuristics may be used in such scenarios as a secondary optimization based on the primary scheduling procedure. Coordinated and randomized microscheduling approaches simulate different strategies for the whole job queue scheduling in order to select the most resource-efficient scenarios. Based on a conservative backfilling scheduling procedure we study how different resources allocation heuristics affect integral job-flow scheduling characteristics in a dedicated simulation environment.