Interprocessor communication with memory constraints
Many parallel applications require periodic redistribution of workloads and associated data. In a distributed memory computer, this redistribution can be difficult if limited memory is available for receiving messages. The authors propose a model for optimizing the exchange of messages under such circumstances which they call the minimum phase remapping problem. They first show that the problem is NP-Complete, and then analyze several methodologies for addressing it. First, they show how the problem can be phrased as an instance of multi-commodity flow. Next, they study a continuous approximation to the problem. They show that this continuous approximation has a solution which requires at most two more phases than the optimal discrete solution, but the question of how to consistently obtain a good discrete solution from the continuous problem remains open. Finally, they devise a simple and practical approximation algorithm for the problem with a bound of 1.5 times the optimal number of phases.
- Research Organization:
- Sandia National Lab. (SNL-NM), Albuquerque, NM (United States); Sandia National Lab. (SNL-CA), Livermore, CA (United States)
- Sponsoring Organization:
- US Department of Energy (US)
- DOE Contract Number:
- AC04-94AL85000
- OSTI ID:
- 756410
- Report Number(s):
- SAND2000-1361C; TRN: AH200022%%164
- Resource Relation:
- Conference: Symposium on Parallel Algorithms and Architectures, Bar Harbor, ME (US), 07/10/2000; Other Information: PBD: 30 May 2000
- Country of Publication:
- United States
- Language:
- English
Similar Records
Quantum computational phase transition in combinatorial problems
On the memory attribution problem: A solution and case study using MPI