Workflow migration in uncertain edge computing environments based on interval many-objective evolutionary algorithm

In edge computing (EC), when the edge server (ES) is processing tasks delivered by the mobile devices (MDs), the MDs move outside the coverage of the ES, where task migration is required to ensure service continuity. Most current research on task migration ignores inter-task dependencies and uncertain computing environments, and it focuses mainly on migration scenarios where MDs have a one-to-one or many-to-one relationship with ESs. Aiming at the problem of workflow migration with multi-MDs and multi-ESs in uncertain environments, this paper proposes an interval many-objective optimized workflow migration in uncertain environments (IMaOWMUE) model that considers transforming uncertainty factors into interval parameters for processing, along with the migration delay, maximum completion time, energy consumption, and load balancing as an objective function, and at the same time, utilize real-time priority scheduling strategies to achieve the fast response of the tasks. Considering the dependency of tasks and the changing characteristics of ES load in a migration environment, this paper designs a migration-based interval many-objective evolutionary algorithm (MI-MaOEA), which adopts an interval confidence strategy to improve algorithm convergence and formulates an objective value-dominated hierarchical sorting and dual-migration selection strategy based on the migration delay and the success rate of the migration to improve the diversity of the populations. Simulation results show that MIMaOEA optimizes 27%, 35%, 14%, and 80% in solving the four objective values of I-MaOWMUE, and enables the solution to have faster converse speed and better distribution.

Automated summary

This paper proposes a model for multi-objective optimized workflow migration in uncertain environments, considering the task dependencies and changing characteristics of edge server load. It also designs a migration-based interval many-objective evolutionary algorithm to solve this problem. Simulation results show that the algorithm achieves significant optimization effects in solving the objective values.