On Tue, 02 Jun 2009 10:06:30 +0200, Kristian Nielsen <[email protected]> wrote: > Alex Yurchenko <[email protected]> writes: > >> I wonder how practicable such setup really is. If one has disk-bound >> application, would not he rather use RAID or something to speed up >> individual writes than trying to execute writes in parallel hoping that >> they get to different drives? I mean acceleration of writes on low level >> always works, parallelization would require suitable application profile >> and careful distribution of the database on multiple drives. Or are we >> talking about multiple RAIDs? > > > RAID == multiple disks. (You can use multiple tablespaces as well, but RAID > (or maybe LVM) is probably the most often used method).
Yes, RAID == multiple disks. But semantics is considerably different. Instead of applying in parallel several potentially conflicting transactions, RAID parallelizes applying of an individual transaction. So it is pure speed-up without side effects. It also has an advantage of more even distribution of data between the drives, so it is more likely to utilize the whole throughput of an array. > And it is not really the writes that are the problems. The problem is the > reads that are needed to figure out which writes to do. (Writes are less of > a > problem, as they can be done asynchroneously). > > And RAID does _not_ speed up an individual read! It only allows to do more > reads in parallel, increasing overall throughput. Hm, I was missing that. I guess you're right here: the only way to speed up the reading part is to do parallel applying. Thanks Alex _______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
