The Principles of Fault-Tolerant and Efficient Parallel Computation
Abstract
The high-performance potential of parallel and distributed computation can only be realized with significant computation speed-ups from the coordinated action of many processors. A basic problem that has to be addressed, in order to realize this potential, is the unreliability of the resulting (highly complex) systems of many processors. The research of N00014-91-J-1613 'Principles of Fault-Tolerant and Efficient Parallel Computation bas focused, primarily. on the algorithmic Principles of fault-tolerant and efficient parallel computing. The desirable combination of reliability and performance is nontrivial since efficiency implies removing redundancy, whereas fault-tolerance requires adding some redundancy to computations
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 15, 1994
- Accession Number
- ADA283254
Entities
People
- Paris O. Kanellakis
Organizations
- Brown University