Paging tradeoffs in distributed-shared-memory multiprocessors

Massively parallel processors have begun using commodity operating systems that support demand paged virtual memory. To evaluate the utility of virtual memory, we measured the behavior of seven shared memory parallel application programs on a simulated distributed shared memory machine. Our results: confirm the importance of gang CPU scheduling; show that a page faulting processor should spin rather than invoice a parallel context switch; show that our parallel programs frequently touch most of their data; and indicate that memory, not just CPUs, must be "gang scheduled". Overall, our experiments demonstrate that demand paging has limited value on current parallel machines because of the applications' synchronization and memory reference patterns and the machines' high page fault and parallel context switch overheads.<<ETX>>

[1]  J. Ousterhout,et al.  Medusa: an experiment in distributed operating system structure , 1980, CACM.

[2]  Peter J. Denning,et al.  The working set model for program behavior , 1968, CACM.

[3]  Harold S. Stone,et al.  Footprints in the cache , 1986, SIGMETRICS '86/PERFORMANCE '86.

[4]  Harold S. Stone,et al.  Footprints in the cache , 1987, TOCS.

[5]  Edward D. Lazowska,et al.  Speedup Versus Efficiency in Parallel Systems , 1989, IEEE Trans. Computers.

[6]  Mary K. Vernon,et al.  Issues in multiprogrammed multiprocessor scheduling , 1990 .

[7]  David H. Bailey,et al.  The Nas Parallel Benchmarks , 1991, Int. J. High Perform. Comput. Appl..

[8]  Larry Rudolph,et al.  Gang Scheduling Performance Benefits for Fine-Grain Synchronization , 1992, J. Parallel Distributed Comput..

[9]  Anoop Gupta,et al.  SPLASH: Stanford parallel applications for shared-memory , 1992, CARN.

[10]  Jack Dongarra,et al.  Integrated Pvm Framework Supports Heterogeneous Network Computing , 1993 .

[11]  James R. Larus,et al.  The Wisconsin Wind Tunnel: virtual prototyping of parallel computers , 1993, SIGMETRICS '93.

[12]  L. Iftode,et al.  Memory servers for multicomputers , 1993, Digest of Papers. Compcon Spring.

[13]  S. T. Leutenegger,et al.  Distributed computing feasibility in a non-dedicated homogeneous distributed system , 1993, Supercomputing '93.

[14]  Jingwen Wang,et al.  Utopia: A load sharing facility for large, heterogeneous distributed computer systems , 1993, Softw. Pract. Exp..

[15]  David A. Patterson,et al.  A case for networks of workstations (now) , 1994, Symposium Record Hot Interconnects II.

[16]  David E. Culler,et al.  A case for NOW (networks of workstation) , 1995, PODC '95.