MIN-Max-Min: A Heuristic Scheduling Algorithm for Jobs across Geo-Distributed Datacenters

In geo-distributed datacenters, tasks in a job often need to run on different sites due to the input data locality or special preference for resources. The completion time (make span) of the job depends on the execution of the slowest task. Considering the heterogeneity of resources and potential skews in number of tasks per job, how to reduce the make span to improve application performance remains an open problem. In this paper, we propose a heuristic scheduling algorithm called MIN-Max-Min that coordinate job scheduling across datacenters. MIN-Max-Min gives priority to select the job with the shortest expected completion time to execute by heuristic rule. Experiments show that compared with first come first service strategy, MIN-Max-Min can reduce the average make span of jobs up to 40% under the simulation load.