Serves a very good purpose in an OLTP environment. It's entirely useless for OLAP. Disk backed tables help, but you're still stuck with in-memory indexes. Pull the plug on the rack, cluster takes a looong time to come back.
Data retention, try purging a few thousand rows, you're limited by MaxNoOfConcurrentOperations. Allocating an object for each row to be inserted/deleted/returned is an impossibility for an OLAP system. The widely accepted solution is to chunk deletes (delete from [table] where date < [target] limit [N])
Quite unfortunate due to possiblity of parallelization of large GROUP BYs and other misc data aggregation.
Maybe in 6.0?
Tuesday, August 12, 2008
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment