Scalable and adaptive resource sharing in PVM
PVM uses round-robin as its default policy for process allocation to processors. The main drawbacks of this policy are the fact that PVM ignores load variations among different nodes and also the inability of PVM to distinguish between machines of different speeds. To redress this deficiency a Resource Manager (RM) is implemented which replaces round-robin with a scalable and adaptive algorithm for resource sharing  providing a High Performance Computing Cluster (HPCC). In this paper an implementation of a Resource Manager is proposed. The RM can be transparently plugged into PVM to offer improved performance to its users. The design of a resource manager to extend PVM is described. A prototype implementation in PVM is then measured to illustrate the utility of the approach. Finally, performance results favorably comparing the enhanced version to the original PVM are presented ...
Unable to display preview. Download preview PDF.
- 1.A.Barak, A. Braverman, I.Gilderman, O. Laadan, ”Performance of PVM and the MOSIX Preemptive Process migration Scheme”, Proc. 7th Israeli Conf. on Computer Systems and Software Engineering, IEEE Computer Society Press 1996.Google Scholar
- 2.A. Barak, S. Guday and R.G. Wheeler. ”The MOSIX Distributed Operating System, Load Balancing for UNIX”, Lecture Notes in Computer Science, Vol. 672, Springer-Verlag, 1993.Google Scholar
- 3.A. Beguelin, J. Dongarra, G.A. Geist, W. Jiang, R. Manchek, V. Sunderam, ”PVM: Parallel Virtual Mchine, a Users Guide and Tutorial for Networked Parallel Computing, MIT Press, Cambridge, MA, 1994.Google Scholar
- 4.S. Crane, K. Twidle, ”Constructing Distributed Unix Utilities in Regis”, Proceedings of the Second International Workshop on Configurable Distributed Systems, March 1994.Google Scholar
- 5.A. Dupuy, J. Schwartz, ”Nest: Network Simulation Tool”, Technical Report, Communications of the ACM, October, No. 10, 1990, Vol. 33.Google Scholar
- 6.Eager D. L., E. D. Lazowska, J. Zahorjan, ”Adaptive Load Sharing in Homogeneous Distributed Systems”, IEEE Trans. on Software Eng., 12(5), pp. 662–675, May 1986.Google Scholar
- 7.G.A. Geist, J.A.Kohl, P.M. Papadopoulos, S.L. Scott ”Beyond PVM 3.4: What we have Learned, Whats Next, and Why”, Oak ridge National Laboratory, Computer Science and Mathematics Division, Oak Ridge, URL: http://www.epm.ornl.gov/pvm/context.ps, 1997Google Scholar
- 8.G.A. Geist, J.A. Kohl, R. Manchek, P. M. Papadopoulus, ”New Features of PVM 3.4”, 1995 EuroPVM Users Group Meeting, Lyon, France, September 1995.Google Scholar
- 9.G.A. Geist, J.A. Kohl, P. M. Papadopoulus, ”CUMULVS”: Providing Fault Tolerance, Visualization and Steering of Parallel Applications”, SIAM, August 1996.Google Scholar
- 10.M. Kapelevich, O. Kremien, ”Scalable Resource Scheduling: Design, Assessment, Prototyping”, Proc. 8th Israeli Conf. on Computer Systems and Software Engineering, IEEE Computer Society Press 1997.Google Scholar
- 11.O. Kremien, J. Kramer, J. Magee, ”Scalable, Adaptive Load Sharing Algorithms”, IEEE Parallel and Distributed Technology, August 1993, 62–70.Google Scholar
- 14.Shivaratri N., P. Krueger, M. Singhal, ”Load Distributing for Locally Distributed Systems”, Computer, 33–44, December 1992.Google Scholar