PGGA: A predictable and grouped genetic algorithm for job scheduling

Abstract

This paper presents a predictable and grouped genetic algorithm (PGGA) for job scheduling. The novelty of the PGGA is twofold: (1) a job workload estimation algorithm is designed to estimate a job workload based on its historical execution records, (2) the divisible load theory (DLT) is employed to predict an optimal fitness value by which the PGGA speeds up the convergence process in searching a large scheduling space. Comparison with traditional scheduling methods such as first-come-first-serve (FCFS) and random scheduling, heuristics such as a typical genetic algorithm, Min-Min and Max-Min indicates that the PGGA is more effective and efficient in finding optimal scheduling solutions.