Author: Robert Hyatt
Date: 22:37:44 11/10/04
Go up one level in this thread
On November 10, 2004 at 11:35:39, Dan Andersson wrote: > The real strength of Tuple spaces is that it can abstract away the underlying >computer architecture. One problem with Tuple spaces is that it introduces a >shared memory to any algorithm. > >MvH Dan Andersson I'm not sure what that means. tuple-space is not a "shared memory" facility at all. A "tuple" is a pair <key><data> that exists in cyberspace somewhere, usually distributed on various nodes in a cluster. You "in" them to read them while leaving them in cyberspace, you "get" them to gain exclusive access and remove them from general availability, and you "out" them to put them out where other processes can see them. The "get" type operation is a mutual-exclusion mechanism that guarantees that only one app can access the tuple at a time. So I guess I miss the concept of "shared memory". I've used C-linda here quite a bit, and we have our own implementation of a tuple-space system that we also use on our clusters... There's nothing shared at all, it is all based on message-passing across tcp/ip sockets...
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.