Author: Vincent Vega
Date: 15:53:56 09/21/00
Go up one level in this thread
On September 21, 2000 at 13:08:11, Ratko V Tomic wrote: >> 2. Send the same data to two or more participants and verify that >> their results match. > >You would need at least 3 nodes working on the same task to make the majority >vote reliable. That factor is in addition to the coordination losses inherent in >the parallelism. True. That factor could be reduced somewhat by a bit more complex schemes like assigning trust values to machines (new ones being least trusted) and by sending test data for which we know already results.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.