An Assessment of Heuristics for Fast Scheduling of Grid Jobs

Due to the dynamic nature of the grid and the frequent arrival of new jobs, rescheduling of already planned and new jobs is a permanent process that is in need of good and fast planning algorithms. This paper extends previous work and deals with newly implemented heuristics for our Global Optimizing Resource Broker and Allocator GORBA. Of a range of possibly usable heuristics, the most promising ones have been chosen for implementation and evaluation. They serve for the following two purposes: Firstly, the heuristics are used to quickly generate feasible schedules. Secondly, these schedules go into the start population of a subsequent run of our Evolutionary Algorithm incorporated in GORBA for improvement. The effect of the selected heuristics is compared to our best simple one used in the first version of GORBA. The investigation is based on two synthetically generated benchmarks representing a load of 300 grid jobs each. A formal definition of the scheduling problem is given together with an assessment of its complexity. The results of the evaluation underline the described intricacy of the problem, because none of the heuristics performs better than our simple one, although they work well on other presumably easier problems.

[1]  Türkay Dereli,et al.  MULTIPLE DISPATCHING RULE BASED HEURISTIC FOR MULTI-OBJECTIVE SCHEDULING OF JOB SHOPS USING TABU SEARCH , 2002 .

[2]  Wilfried Jakob,et al.  Construction of Benchmarks for Comparison of Grid Resource Planning Algorithms , 2018, ICSOFT.

[3]  Wilfried Jakob,et al.  Fast multi-objective rescheduling of grid jobs to constrained resources by heuristics and evolution. , 2010 .

[4]  L. Goddard,et al.  Operations Research (OR) , 2007 .

[5]  Wilfried Jakob,et al.  Fast Multi-objective Scheduling of Jobs to Constrained Resources Using a Hybrid Evolutionary Algorithm , 2008, PPSN.

[6]  Stéphane Dauzère-Pérès,et al.  A modified shifting bottleneck procedure for job-shop scheduling , 1993 .

[7]  Egon Balas,et al.  The Shifting Bottleneck Procedure for Job Shop Scheduling , 1988 .

[8]  G. Thompson,et al.  Algorithms for Solving Production-Scheduling Problems , 1960 .

[9]  Christian Blume.  GLEAM - A System for Simulated 'Intuitive Learning' , 1990, PPSN.

[10]  Wilfried Jakob,et al.  Towards an Adaptive Multimeme Algorithm for Parameter Optimisation Suiting the Engineers' Needs , 2006, PPSN.

[11]  P. Simin Pulat,et al.  The shifting bottleneck procedure for job-shops with parallel machines , 2006 .

[12]  Mauricio G. C. Resende,et al.  Greedy Randomized Adaptive Search Procedures , 1995, J. Glob. Optim..

[13]  Christian Blume,et al.  GLEAM - An Evolutionary Algorithm for Planning and Control Based on Evolution Strategy , 2002, GECCO Late Breaking Papers.

[14]  Christian Blume,et al.  GLEAM - General Learning Evolutionary Algorithm and Method : ein Evolutionärer Algorithmus und seine Anwendungen , 2009 .

[15]  Peter Brucker,et al.  Scheduling Algorithms , 1995 .

[16]  Anne Setämaa-Kärkkäinen,et al.  Best compromise solution for a new multiobjective scheduling problem , 2006, Comput. Oper. Res..

[17]  Christoph J. Schuster.  No-wait Job-Shop Scheduling: Komplexität und Local Search , 2003 .

[18]  Peter Brucker,et al.  Complex Scheduling , 2006 .