Обсуждение: Multi DB performance test
"Tadej Kanizar" <Tadej.Kanizar@sagit-server.com> writes: > http://devloop.org.uk/documentation/database-performance/ > Any comments? Old news, see this thread: http://archives.postgresql.org/pgsql-general/2007-02/msg00806.php Most of us stopped reading about the place where they explained that (a) they did zero tuning on each database, and (b) they were comparing myisam to transactional engines. It's amazing how many people do not understand that database performance testing is actually a pretty hard problem in itself --- simplistic tests will almost never show you the real-world bottlenecks. regards, tom lane
On 12.03.2007, at 22:46, Tom Lane wrote: > Most of us stopped reading about the place where they explained > that (a) they did zero tuning on each database, and (b) they were > comparing myisam to transactional engines. What I don't like about things like that, is that some morons throw something together they call a "performance test", add some colorful graphs to it and it will be quoted for the next ten years, "decision makers" will base their recommendations on it and a lot of followers will just believe. Whether or not this applies to the referenced test - judge yourself. cug
On Mon, 2007-03-12 at 23:46, Tom Lane wrote: > "Tadej Kanizar" <Tadej.Kanizar@sagit-server.com> writes: > > http://devloop.org.uk/documentation/database-performance/ > > Any comments? > > Old news, see this thread: > http://archives.postgresql.org/pgsql-general/2007-02/msg00806.php > > Most of us stopped reading about the place where they explained > that (a) they did zero tuning on each database, and (b) they were > comparing myisam to transactional engines. > > It's amazing how many people do not understand that database performance > testing is actually a pretty hard problem in itself --- simplistic tests > will almost never show you the real-world bottlenecks. While I still agree that the tests are overly simple, and don't really give a good feel for a real world database, I just want to point out that they have updated their testing and results to now include innodb and berkely table types for mysql. Sadly, they still do no transactions, they just compare single statement innodb / berkely to myisam.