Parallel task graph scheduling has long been an important research topic in the field of parallel processing and is well known to be a challenging NP-complete problem. With the advancement of technology and emergence of grid and cloud computing, now many large-scale scientific and engineering applications are usually constructed as workflows, which are similar to traditional parallel task graphs in structure, due to large amounts of interrelated computation and communication. Therefore, workflow scheduling, especially for concurrent and online workflows, has become a crucial research issue.
Among various workflow scheduling methods in the literature, listing-based and clustering-based scheduling are the two most important and commonly used categories of approaches. In this thesis, we make two contributions to these two types of workflow scheduling, respectively. In the first part of contribution, we developed new task ranking and allocation methods for single workflow scheduling based on list-based scheduling approaches.
In the second contribution, we proposed efficient task group allocation methods, considering both resource fitness and tasks’ EFT (Earliest Finish Time), for concurrent workflow scheduling using cluster-based scheduling approaches.
The proposed approaches were evaluated with a series of simulation experiments and compared to typical methods, such as the HEFT [16], the lookahead variant of HEFT [5], the pure best-fit approach [13], PCH approach [21][22], and the distributed gap search approach [18]. The experimental results show that our approaches outperform existing methods significantly, achieving up to 11.8% and 15.5% performance improvement in terms of average makespan for list-based scheduling and clustering-based scheduling, respectively.
In the experiments, we found that the effectiveness of some approaches is greatly affected by workflow structure or the system load. For example, the top + bottom task ranking methods is effective for general workflow, but performs poorly for workflows of fork-join structure.
Moreover, the FST task allocation method can bring advantages only under medium or high system load. Further research work is required to find the root causes and clarify what characteristics of workflows and conditions of system load would determine the best choice of task ranking and allocation methods. Such future work will shed light on how to develop a more advanced hybrid approach for further performance improvement of workflow scheduling.
References
[1] A. Mandal, K. Kennedy, C. Koelbel, G. Marin, J. Mellor-Crummey, B. Liu, and L.
Johnsson. “Scheduling Strategies for Mapping Application workflows onto the Grid,”
Proceedings of the 14th IEEE Symposium on High Performance Distributed Computing (HPDC 14), pp. 125-134, 2005.
[2] A. Ramakrishnan, G. Singh, H. Zhao, E. Deelman, R. Sakellariou, K. Vahi, K.
Blackburn, D. Meyers, and M. Samidi, “Scheduling data-intensive workflows onto storage-constrained distributed resources,” in CCGRID ’07:Proceedings of the Seventh IEEE International Symposium on Cluster Computing and the Grid. Washington, DC,USA: IEEE Computer Society, 2007, pp. 401–409.
[3] ASKALON, http://www.dps.uibk.ac.at/projects/teuta/ (2011/12) [4] Business Process Execution Language (BPEL),
http://en.wikipedia.org/wiki/Business_Process_Execution_Language (2013/07)
[5] Ching-Hsien Hsu, Chih-Wei Hsieh, Chao-Tung Yang, “A Generalized Critical Task Anticipation Technique for DAG Scheduling”,Proceedings of ICA3PP 2007, LNCS 4494, pp. 493–505, 2007.
[6] DAGman, http://www.cs.wisc.edu/condor/manual/v7.0/(2011/12)
[7] E. Deelman, G. Singh, and C. Kesselman. “Optimizing Grid-based Workflow Execution,”
Journal of Grid Computing, 3(3): 201-219, 2005.
[8] Fishman, G.S, Principles of discrete event simulation, John Wiley and Sons,New York, NY, 1978 Jan 01.
[9] F. Afrati, C. Papadimitriou, G. Papageorgiou, “Scheduling DAGs to Minimize Time and Communication”, Proceedings of the 3rd Aegean Workshop on Computing: VLSI Algorithms and Architectures, Springer-Verlag, pp. 134–138, 1988.
[10] F. R. L. Cicerre, E. R. M. Madeira, L. E. Buzato, “A Hierarchical Process Execution Support for Grid Computing”, Concurrency and Computation: Practice and Experience, vol.18 no.6, pp. 581-594 2006.
[11] G. C. Sih and E. A. Lee. “A Compile-Time Scheduling Heuristic for Interconnection-Constrained Heterogeneous Processor Architectures,” IEEE Transactions on Parallel and Distributed Systems, vol. 4, no. 2, pp. 175-186, Feb 1993.
[12] G. Park, B. Shirazi, and J. Marquis. “DFRN: A New Approach for Duplication Based Scheduling for Distributed Memory Multi-processor Systems,” Proc. Int'l Conf. Parallel Processing, pp. 157-166, 1997.
[13] G. L. Stavrinides, H. D. Karatza, “Scheduling Multiple Task Graphs in Heterogeneous Distributed Real-Time Systems by Exploiting Schedule Holes with Bin Packing Techniques,” Proc.Simulation Modelling Practice and Theory, vol. 19, issue 1, pp.
540-552, January 2011.
[14] Gridbus,http://www.cloudbus.org/workflow/ (2011/12)
[15] H. EI-Rewini and T. G. Lewis. “Scheduling Parallel Program Tasks onto Arbitrary Target Machines,”Proc.Journal of Parallel and Distributed Computing, vol. 9, pp. 138-153,
and Distributed Systems, 2(13):260-247, 2002.
[17] H. Zhao and R. Sakellarious, “Scheduling Multiple DAGs onto Heterogeneous Systems,”
Proceedings of the 15th Heterogeneous Computing Workshop (HCW), Rhodes Island, Greece, April 2006.
[18] H. J. Jiang, K. C. Huang, H. Y. Chang, D. S. Gu and P. J. Shih, "Scheduling Concurrent Workflows in HPC Cloud Through Exploiting Schedule Gaps,” Proceedings of the 11th International Conference on Algorithms and Architectures for Parallel Processing (ICA3PP 2011),27-28 October 2011.
[19] J. Liou and M.A. Palis, “An Efficient Clustering Heuristic for Scheduling DAGs on Multiprocessors”, Proceedings of the8thSymposium. Parallel and Distributed Processing, 1996.
[20] J. Blythe, S. Jain, E. Deelman, Y. Gil, K. Vahi, A. Mandal, K. Kennedy, “Task Scheduling Strategies for Workflow-based Applications in Grid,” Proceedings of the IEEE International Symposium on Cluster Computing and Grid, pp. 759-767 Vol. 2, 9-12 May 2005.
[21] L. F. Bittencourt and E. R. M. Madeira, “Fulfilling Task Dependence Gaps for Workflow Scheduling on Grids,” Proceedings of the 3rd IEEE International Conference on Signal-Image Technology and Internet Based Systems (SITIS), Dec, 2007.
[22] L. F. Bittencourt, E.R.M. Madeira, “A Performance-Oriented Adaptive Scheduler for Dependent Tasks on Grids,” Proc.Concurrency and Computation: Practice and Experience vol. 20, pp. 1029–1049, 2008.
[23] L. F. Bittencourt and E. R. M. Madeira, “Towards the Scheduling of Multiple Workflows
on Computational Grids,” Proc.Journal of Grid Computing, vo.1, no.8, pp. 419-441, Number 3, 2009
[24] L. F. Bittencourt, RizosSakellariou,Edmundo, R. M. Madeira, “DAG Scheduling Using a Lookahead Variantof the Heterogeneous Earliest Finish Time Algorithm”,Proceedings of the 18th Euromicro Conference on Parallel, Distributed and Network-based Processing,2010.
[25] M.R. Gary and D.S. Johnson,Computers and Intractability A Guide to the Theory of NP-Completeness, W.H. Freeman and Co., San Francisco 1979
[26] M. Wu and D. Gajski. “Hypertool: A Programming Aid for Message Passing Systems,”
Proc.IEEE Transactions on Parallel and Distributed Systems, vo.l,no.1, pp. 330-343, July 1990.
[27] M. Rahman, S. Venugopal, R. Buyya, “A Dynamic Critical Path Algorithm for Scheduling Scientific Workflow Applications on Global Grids”, Proceedings of the 3rd IEEE Int. Conf. on e-Science and Grid Computing, pp. 35–42, 2007.
[28] M. Wieczorek, R. Prodan, A. Hoheisel, M. Wieczorek, R. Prodan, A. Hoheisel,
“Taxonomies of the Multi-Criteria Grid Workflow Scheduling Problem”,Grid middleware and services, pp. 237-264, 2008.
[29] O. Sinnen, Task Scheduling for Parallel Systems, John Wiley, 2007.
[30] Pegasus,http://pegasus.isi.edu/ (2011/12)
[31] S. J. Kim and J.C. Browne, “A General Approach to Mapping of Parallel Computation upon Multiprocessor Architectures”, Proceedings of Int’l Conf. Parallel Processing, vol.
2, pp. 1-8, 1988.
[32] T. Yang and A. Gerasoulis. “DSC: Scheduling Parallel Tasks on an Unbounded Number of Processors,” Proc.IEEE Transactions on Parallel and Distributed Systems, vol. 5, no.
9, pp. 951-967, Sept. 1994.
[33] T. D. Braun, H. J. Siegel, N. Beck, L. L. Bölöni, M. Maheswaran, A. I. Reuther, J.
P. Robertson, M. D. Theys, B. Yao, D. Hensgen, R. F. Freund, “A Comparison of Eleven Static Heuristics for Mapping a Class of Independent Tasks onto Heterogeneous Distributed Computing Systems,”Proc.Journal of Parallel and Distributed Computing, vol.61, no.6, June 2001 , pp. 810-837(28).
[34] T. N'takpe' and F. Suter, Concurrent Scheduling of Parallel Task Graphs on Multi-Clusters Using Constrained Resource Allocations, Rapport de recherché n° 6774, December 2008.
[35] Y. Kwok and I. Ahmad. “Dynamic Critical-Path Scheduling: An Effective Technique for Allocating Task Graphs to Multi-processors,” Proc.IEEE Transactions on Parallel and Distributed Systems, vol. 7, no. 5, pp. 506-521, May 1996.
[36] Y. Gu and Q. Wu, “Optimizing Distributed Computing Workflows in Heterogeneous Network Environments”, Proceedings of the 11th International Conference on Distributed Computing and Networking, pp. 142-154, Kolkata, India, January 03-06, 2010.
[37] Y. L. Tsai, K. C. Huang, H. Y. Chang, J. Ko, E. T. Wang, C. H. Hsu, “Scheduling
Multiple Scientific and Engineering Workflows through Task Clustering and Best-Fit Allocation”, Proceedings of 2012 IEEE Eighth World Congress on Services ,June 24- 29,2012
[38] Z. Yu and W. Shi, “A Planner-Guided Scheduling Strategy for Multiple Workflow Applications,” Proceedings of the 37th International Conference on Parallel Processing Workshops, ICPPW 08, 8-12 Sept. 2008.