Talk:Distributed shared memory

Earlier comment

how about numa, and the Altix? it does fall in shared memory category doesn't it?! — Preceding unsigned comment added by 212.138.64.171 (talk) 19:27, 21 March 2005 (UTC)[reply]

NUMA is not DSM

NUMA and DSM are distinct. With NUMA a node can access memory elsewhere albeit less quickly than local memory. With DSM a node cannot directly access memory elsewhere. To access another node's memory it must engage the other node's cooperation (e.g., via messages). To ensure that all nodes have a coherent view of memory, now that's where things get interesting ... — Preceding unsigned comment added by 65.96.114.69 (talk) 00:18, 7 June 2005 (UTC) ‎[reply]

The 2011 edition of Hennessy and Patterson, Computer architecture : a quantitative approach says on page 348 that "A DSM [Distributed Shared Memory] multiprocessor is also called a NUMA (nonuniform memory access), since the access time depends on the location of a data word in memory." So perhaps our article needs to be updated or changed to agree with modern reliable sources. JonH (talk) 14:42, 31 July 2013 (UTC)[reply]

Delphi?

The "Delphi" link goes to Delphi programming language; the linked article mentions nothing about distributed shared memory. This leads me to believe the Delphi that's an example of DSM is not the same as the Delphi programming language. A Google search turns up this paper, which is about a distributed shared memory system called Delphi and doesn't mention Borland afaict. Anyone know what's going on?

Cjoev 01:22, 4 January 2006 (UTC)[reply]

The link is broken. I intended to fix it, but I forgot about it. Of course it should link to new entry about DSM Delphi system, and NOT about Borland Delphi.

Szopen 09:20, 4 January 2006 (UTC)[reply]

MOSIX

Mosix does not support DSM, it migrates memory from node to node, but a process can't access memory of any other node beside its hosting one. —Preceding unsigned comment added by 132.65.96.97 (talk) 07:21, 25 January 2011 (UTC)[reply]