Task Allocation in Distributed Mesh-Connected Machine Learning System: Simplified Busy List Algorithm with Q-Learning Based Queuing

In the era where organizations gather and process more and more data, Machine Learning (ML) techniques become increasingly important. Considering “Big Data,” ML usually involves intensive data processing and high-performance computing. To meet the growing requirements, efficient distributed and parallel systems are key factors. In this paper, we consider mesh-based distributed system and task allocation methods in the system. We focus especially on the impact of intelligent queuing in task allocation algorithms. A new SBL algorithm with Q-Learning queuing is presented. In addition, the new SBL technique is compared to other well-known allocation schemes, which are discussed as well. The comparison is made using an implemented experimentation system and simulation results are presented. The results confirm that SBL algorithm and the queuing system deliver good performance characteristic.

[1]  Tadeusz Luba,et al.  Energy characteristic of a processor allocator and a network-on-chip , 2011, Int. J. Appl. Math. Comput. Sci..

[2]  Vincenzo Cutello,et al.  Parallel Problem Solving from Nature - PPSN XII , 2012, Lecture Notes in Computer Science.

[3]  Hee Yong Youn,et al.  An Efficient Task Allocation Scheme for 2D Mesh Architectures , 1997, IEEE Trans. Parallel Distributed Syst..

[4]  Nian-Feng Tzeng,et al.  An efficient submesh allocation strategy for mesh computer systems , 1991, [1991] Proceedings. 11th International Conference on Distributed Computing Systems.

[5]  Leszek Koszalka,et al.  Comparison of SBA - Family Task Allocation Algorithms for Mesh Structured Networks , 2006, ISPA Workshops.

[6]  Leszek Koszalka,et al.  Evaluation Scheme for NoC based CMP with Integrated Processor Management System , 2010 .

[7]  Geyong Min.  Frontiers of High Performance Computing and Networking - ISPA 2006 Workshops, ISPA 2006 International Workshops, FHPCN, XHPC, S-GRACE, GridGIS, HPC-GTP, PDCE, ParDMCom, WOMP, ISDF, and UPWN, Sorrento, Italy, December 4-7, 2006, Proceedings , 2006, ISPA Workshops.

[8]  Fan Wu,et al.  Processor Allocation in the Mesh Multiprocessors Using the Leapfrog Method , 2003, IEEE Trans. Parallel Distributed Syst..

[9]  Kenneth A. De Jong,et al.  A Spatial EA Framework for Parallelizing Machine Learning Methods , 2012, PPSN.

[10]  Chita R. Das,et al.  A Fast and Efficient Processor Allocation Scheme for Mesh-Connected Multicomputers , 2002, IEEE Trans. Computers.

[11]  Sujatha R. Upadhyaya,et al.  Parallel approaches to machine learning - A comprehensive survey , 2013, J. Parallel Distributed Comput..

[12]  Laxmi N. Bhuyan,et al.  An Adaptive Submesh Allocation Strategy for Two-Dimensional Mesh Connected Systems , 1993, 1993 International Conference on Parallel Processing - ICPP'93.

[13]  Yahui Zhu,et al.  Efficient Processor Allocation Strategie for Mesh-Connected Parallel Computers , 1992, J. Parallel Distributed Comput..

[14]  Neoklis Polyzotis,et al.  Declarative Systems for Large-Scale Machine Learning , 2012, IEEE Data Eng. Bull..

[15]  Henry Selvaraj,et al.  Fast and efficient processor allocation algorithm for torus-based chip multiprocessors , 2011, Comput. Electr. Eng..