Parallel Workstation Cluster for Algorithm Development, Parameter Scoping, and Data Analysis
Abstract
A cluster of workstations is connected with a fast network and used as a local parallel computer. This approach mimics the interprocessor communication and scalability of the large machines at a fraction of the cost. This system uses off-the-shelf hardware and public domain software, which makes the system simple to assemble and maintain. The parallel workstation cluster has already significantly contributed to our AFOSR project. The cluster accelerates the algorithm development process because it eliminates the delays associated with remote computing on shared resources. The parallel workstation cluster performs simulations to scope a parameter space for optimization and design. A parallel supercomputer is then used for the detailed scoping using fewer runs. This approach has helped to alleviate the over-subscription of the shared resources and has allowed us to obtain the computational results in less time.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 26, 1999
- Accession Number
- ADA363032
Entities
People
- Uri Shumlak
Organizations
- University of Washington