Fault Tolerant Design for Multistage Routing Networks
Abstract
As the size of digital systems increases, the average length of time between single component failures diminishes. To avoid component related failures, large computers must be fault-tolerant; that is, the computer must perform correctly even when some components fail. This paper concentrates on providing fault-tolerance in the interconnection network for massively parallel MIMD computers. Particularly, the focus is on methods for achieving a high degree of fault-tolerance in multistage routing networks. A multipath scheme is described for providing end-to-end fault-tolerance on large networks. The scheme improves routing performance while keeping network latency low. The novel routing component RN1 is described which implements this scheme, showing how it can be the basic building block for fault-tolerant multistage routing networks. (rh)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1990
- Accession Number
- ADA224267
Entities
People
- André DeHon
- Henry Minsky
- Tom Knight
Organizations
- Massachusetts Institute of Technology