A Probabilistic Evolutionary Hybrid (EDA-GA) for Multi-Objective Task Scheduling in Cloud Computing

Abstract

Cloud computing requires effective task scheduling to allocate heterogeneous workloads on virtual machines while maintaining quality of service and resource utilisation. The study aims to solve the problem of minimising task completion time and improving load balance simultaneously in cloud environments. To tackle this problem, the EDA-GA hybrid algorithm is proposed which is composed of the probabilistic solution generation of Estimation of Distribution Algorithm and the crossover and mutation operations of Genetic Algorithm. The experiment was implemented in a simulation environment based on CloudSim with synthetic CSV datasets containing parameters of cloudlets, virtual machine configurations and scheduling scenarios. The proposed method was compared with standalone GA and standalone EDA based on makespan, load imbalance and optimisation time. Results show that EDA-GA outperformed EDA, reducing the average makespan from 93.7858 to 90.7256 and the average load imbalance from 0.2050 to 0.1287. However, the best overall average makespan and load balance was achieved by GA. This study introduces a hybrid scheduling framework and highlights its advantages, shortfalls and future optimisation requirements.