All these grid, clustering talk is very interesting. Please continue.
:) On Jun 17, 2007, at ,Jun 17, 3:28PM, Orlando Andico wrote: > coherence can be compared most directly to Terracotta. > > however Terracotta only provides 2-way clustering, and the second node > must be a standby. coherence provides a single-image shared memory for > multiple java applications across a grid; and all grid nodes are fully > active (can read/write to the shared memory) and replicated. so losing > a node is not fatal (all data on a node is replicated on AT LEAST one > other node). > > adding new nodes means some backups get copied to that new node to > distribute the load more effectively. losing nodes means the data on > that node is shifted to the backup node, while backups on that node > are re-copied to one of the surviving nodes. > > with a 1000-node cluster, coherence can do ~25M aggregations per > second (that's 25M puts into the shared store). of course reads are > much, much faster, and a general workload would have an app reading > stuff from the store, crunching on it for a long-ish time, and then > doing a put back into the store once done. > > > -- Garibaldi V. Melecio <[EMAIL PROTECTED]> +63.917.951.4054 - _________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List [email protected] (#PLUG @ irc.free.net.ph) Read the Guidelines: http://linux.org.ph/lists Searchable Archives: http://archives.free.net.ph
