Scheduling Bag-of-Tasks with Selective Imprecise Computations in a Mist - Fog - Cloud Environment

This paper investigates the scheduling of bag-of-task (BoT) jobs within a three-tier architecture comprising mist, fog, and cloud layers, each characterized by heterogeneous computing capacities. Specifically, mist resources are deemed optimal for processing tasks of BoTs with stringent security requirements, referred to as “mist jobs”, while BoTs with lesser security demands are categorized as “fog jobs” and can be assigned to either fog or mist resources. Conversely, BoTs lacking security prerequisites, denoted as “cloud jobs” are suitable for processing on cloud, fog, or mist resources. To expedite the execution of mist and fog jobs, a method employing imprecise computations for cloud job tasks under specific conditions is employed. With simulation experiments, this study offers valuable insights into the impact of these approximations on system performance, based on the number of tasks within cloud jobs and their execution layer (mist or fog), accompanied by a small decrease in result precision. Various system load levels and workload types are studied to provide a comprehensive analysis.