You might be interested in this presentation from Linux World Conference August, 2002, which gave a comparison of Linux Journaling file system types based on io tests run at the Open Source Development Lab:
http://www.osdl.org/presentations/lwe-jgfs.pdf On Wed, 2003-01-22 at 07:01, Michael T. Babcock wrote: > David Brodbeck wrote: > > >If Nicholas does this, I hope he'll report back here with what he finds out. > > > > I would recommend that as well. > > >I'm facing the same choice very soon, for a database that will eventually have >millions of entries (but each individual entry very small.) I'm trying to decide >whether to go with reiserfs or ext3. ext3's my current favorite, but only because >it's more of a known, stable quantity. > > > > If I may cause a flame-war here (j/k), reiserfs is a very known, very > stable quantity. I've been using reiserfs for years now and it has > never failed me except in the case of a physical drive failure (and it > still saved me most of my files). Reiserfs is _very_ fast compared with > the ext filesystems; I can't imagine running my /home and /var > partitions on anything else ever again (and the rest are ext3 only > because the RedHat installer wouldn't do reiserfs during the install). > > For what its worth, reiser4 (the new version of reiserfs that seems > almost complete) as well as a patched version of reiserfs (version 3) > have hooks in them to allow software (currently Squid) to use the > internal tree and storage routines and bypass the filesystem layer > altogether. This would make a lot of sense for a program like the mysql > daemon (much as InnoDB probably does for its raw storage already) and > I'd be very interested in the speed with which InnoDB can open and find > records vs. the tree structure speeds in reiser4. Reiser4 is also > supposed to have full filesystem transaction support, which is a great > feature for databases in the first place. > > If you want an idea of how ReiserFS is built to work, you'd probably > (untested) get quite good performance (for a specific group of test > cases) out of writing a MySQL backend that stored tables as directories > under their database directories with row values as files under row > directories, with directories named by their primary index values > (randomly otherwise?). It would make an interesting project at any rate. -- Mary Edie Meredith <[EMAIL PROTECTED]> Open Source Development Lab --------------------------------------------------------------------- Before posting, please check: http://www.mysql.com/manual.php (the manual) http://lists.mysql.com/ (the list archive) To request this thread, e-mail <[EMAIL PROTECTED]> To unsubscribe, e-mail <[EMAIL PROTECTED]> Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
