At the 2004 Users Conference in Orlando in April there were two sessions on optimizing MySQL hosted by a MySQL staffer who's name eludes me for the moment. He told the assembled masses that in benchmarks he ran that innodb_file_per_table was somewhat faster than using the large innodb table space. I didn't get the impression it was like 50% faster or anything, but once finished optimizing indexes and so on any gains are likely to be in small pieces, but they all add up.
I can see the logic of it of course... Most of our servers are running 40Gbyte InnoDB table spaces, two are running 100G space. Some of our tables are small, some have only 7 rows of 2 columns... It must be easier for InnoDB to find 100bytes of data in its own file rather than in 100GBytes of shared table space. So I donšt have anything quantitative, just hearsay from the folks at MySQL who are the performance and fine tuning experts. As to it's being "new"... It's different. It's as new as MySQL 4.1 - if you are using 4.1 then it's no newer than anything else there. MySQL staffers were giving us benchmarks with it back in April at the Users Conference, and I had already figured out I wanted to do it last January, I've just been waiting for a) the "production" version, and b) an opportunity to down my services. MySQL historically has released very stable products by the time they get to "Beta". We always wait till it goes "Production" because we couldn't explain to a client why a problem occurred on beta software, but it's only labeling. MySQL beta typically is more stable than most folks release.1 or release.2 versions. This is because of the very large base of people around the globe using and testing MySQL and contributing to it's development. Best Regards, Bruce On 1/6/05 7:38 AM, "Ken Menzel" <[EMAIL PROTECTED]> wrote: > Hi Bruce > > <SNIP> > ----- Original Message ----- > From: "Bruce Dembecki" <[EMAIL PROTECTED]> > To: <mysql@lists.mysql.com> > Sent: Thursday, December 30, 2004 2:51 AM > Subject: Re: Fixing "the worst InnoDB corruption bug in 3 years" - > when >> >> As a side note with demonstrated performance increases when using >> innodb_file_per_table why aren't more people using it? >> >> Best Regards, Bruce > </SNIP> > > What demonstrated performance increases are you referring to? I would > love to use file_per_table, but as it is new we are very conservative > with our production DB's and this feature is too new. But if there > are "demonstrated performance increases" we would love to start using > it! > > Thanks, > Ken > -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
