Scheduling Gang Jobs with Partial Computations in a Collaborative Fog-Cloud System

Fog-cloud collaborative computing can effectively meet the diverse execution requirements of complex jobs, such as gang jobs. However, this task involves a challenge in the case where jobs of gang type exhibit average arrival rates which vary over time and distinct security requirements. In such scenarios, under certain conditions, it is necessary to partially execute all parallel component tasks of non-security-sensitive gang jobs to avoid delaying security-sensitive gang jobs, which must be exclusively executed in the fog. This paper investigates various gang scheduling strategies when workloads exhibit this type of characteristics within a collaborative fog-cloud system. Specifically, we examine variations of known gang scheduling strategies, incorporating partial computations and comparing these policies when jobs are fully executed. Our objective is to compare the performance of the scheduling techniques under various conditions of partial computations and varying workload scenarios. The results of the simulations performed indicate that while the scheduling approaches benefit from partial computations, there is a slight decrease in result precision.