On the effectiveness of combining in resolving hot spot'' contention

Journal Article · Tue Feb 01 04:00:00 EST 1994 · Journal of Parallel and Distributed Computing; (United States)

OSTI ID:7279722

Lee, Gyungho ^[1]; Kruskal, C P ^[2]; Kuck, D J ^[3]

---

1. Univ. of Minnesota, Minneapolis, MN (United States). Dept. of Electrical Engineering
2. Univ. of Maryland, College Park, MD (United States). Dept. of Computer Science
3. Univ. of Illinois, Urbana, IL (United States). Center for Supercomputing Research and Development

Sharing the memory in a parallel computer suggests that there is a possibility of many processors requesting a shared variable at virtually the same time. This idea has been formalized in the hot spot'' traffic model, where a fixed fraction of memory requests is for a single shared variable. Under the model, it has been shown that such concurrent requests to shared variable can create contention serious enough to stall large machines. As an effective method of alleviating this type of contention, combining,'' in which several requests for the same variable can be combined into a single request, has been suggested. The NYU Ultra computer and the IBM RP3 machine use pairwise'' combining, in which only two requests for the same variable can be combined at a switch of the processor-memory interconnection. The authors study the effectiveness of combining. In particular, it turns out that pairwise combining is slightly too restrictive to handle hot spots if the machines size becomes very large. They suggest ways to overcome this weakness.

• DOE Contract Number: FG02-85ER25001
• OSTI ID: 7279722
• Journal Information: Journal of Parallel and Distributed Computing; (United States), Journal Name: Journal of Parallel and Distributed Computing; (United States) Vol. 20:2; ISSN JPDCER; ISSN 0743-7315
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200* -- Mathematics & Computers
ARRAY PROCESSORS
COMPUTER ARCHITECTURE
MEMORY MANAGEMENT
PARALLEL PROCESSING
PERFORMANCE
PROGRAMMING