Daniel Kiss schrieb:
However, I'm more interested in practical experience with huge databases. How effective is MySQL (with InnoDB) working with tables containing millions or rather billions of rows? How about the response time of queries, which return a few dozens of rows from these big tables when using two, three or more table joins? What is a suggester hardware when we're expecting the database to grow extremely large? How stable is MySQL and the InnoDB engine with terrabytes behind? Etc, etc, etc...
(Of course we assume that the database structure is very well designed and using optimal indexing.)
We are using a dual-opteron machine with 8gigs of RAM and RAID-10 (3ware escalade controller).
No problems, we are very happy (biggest tables are around 300 million rows). When we set the system up I did a few performance tests and hat 6 clients with each 10 threads running that did random joins over a few tables (yes, the big ones) and fetched a few thousand rows with each query, but this didn't bother the server much. At that point everything was fine for me cause I knew everything would work out for our kind of usage and did no further "benchmarks" or such. This Server currently has around 500 queries/second and top tells me 60% idle for the cpus (debian linux, mysql 4.1, both for amd64)
The whole speed of queries and joins depends on the indexes. It ist not only important that your database structure is well designed but that all indexes fit into the RAM. If mysql (or any other DB) needs to read its indexes from disc for every query because it doesn't fit into memory, the performance will drop dramatically.
I started a thread on this list I think about a year ago(?) about which hardware to use and we ended up with a dual-opteron, tyan-board and 3ware escalade-controller for a sata-raid-10. I was told the Opterons perform better than Xeons.
Jan
-- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
