Techniques for reducing consistency-related communication in distributed shared-memory systems

作者: John B. Carter , John K. Bennett , Willy Zwaenepoel

DOI: 10.1145/210126.210127

关键词:

摘要: Distributed shared memory (DSM) is an abstraction of on a distributed-memory machine. Hardware DSM systems support this at the architecture level; software within runtime system. One key problems in building efficient system to reduce amount communication needed keep distributed memories consistent. In article we present four techniques for doing so: release consistency; multiple consistency protocols; write-shared and update-with-timeout mechanism. These have been implemented Munin We compare performance seven application programs: first their when using message passing, then running conventional that does not embody preceding techniques. On 16-processor cluster workstations, Munin's 5% passing out applications. For other three, 29 33%. Detailed analysis two these three applications indicates addition function-shipping capability would bring 7% message-passing performance. Compared system, achieves improvements ranging from few several hundred percent, depending application.

参考文章(26)
Randy H. Katz, Susan J. Eggers, A Characterization of Sharing In Parallel Programs And Its University of California at Berkeley. ,(1987)
Sandhya Dwarkadas, Alan L. Cox, Willy Zwaenepoel, Pete Keleher, TreadMarks: distributed shared memory on standard workstations and operating systems usenix winter technical conference. pp. 10- 10 ,(1994)
J. M. Bernabéu-Aubán, M. P. Pearson, M. Ahamad, R. J. LeBlanc, P. Dasgupta, C. J. Wilkenloh, M. Y. A. Khalidi, U. Ramachandran, P. W. Hutto, W. F. Appelbe, S. Menon, R. C. Chen, R. Ananthanarayanan, The Design and Implementation of the Clouds Distributed Operating System. Computing Systems. ,vol. 3, pp. 11- 46 ,(1989)
Andrew W. Wilson, Richard P. LaRowe, Hiding shared memory reference latency on the Galactica Net distributed shared memory architecture Journal of Parallel and Distributed Computing. ,vol. 15, pp. 351- 367 ,(1992) , 10.1016/0743-7315(92)90049-S
J. Chase, F. Amador, E. Lazowska, H. Levy, R. Littlefield, The Amber system: parallel programming on a network of multiprocessors symposium on operating systems principles. ,vol. 23, pp. 147- 158 ,(1989) , 10.1145/74850.74865
Kai Li, Paul Hudak, Memory coherence in shared virtual memory systems ACM Transactions on Computer Systems. ,vol. 7, pp. 321- 359 ,(1989) , 10.1145/75104.75105
Chandramohan A. Thekkath, Henry M. Levy, Limits to low-latency communication on high-speed networks ACM Transactions on Computer Systems. ,vol. 11, pp. 179- 203 ,(1993) , 10.1145/151244.151247
Lamport, How to Make a Multiprocessor Computer That Correctly Executes Multiprocess Programs IEEE Transactions on Computers. ,vol. 28, pp. 690- 691 ,(1979) , 10.1109/TC.1979.1675439
Anna R. Karlin, Mark S. Manasse, Larry Rudolph, Daniel D. Sleator, Competitive snoopy caching 27th Annual Symposium on Foundations of Computer Science (sfcs 1986). pp. 244- 254 ,(1986) , 10.1109/SFCS.1986.14