I've been doing some benchmarking recently to satisfy the curiosity about 5.1's performance compared with 4.1. Â The major question this time revolves around how much additional performance an external RAID array can provide (for us it's typically beyond the 6 drives a Dell 2950 can hold).Â
These tests are done on using an MSA-30 drive enclosure with 15k-SCSI drives.  The testing framework is sysbench oltp.  The test names are hopefully fairly obvious:  selects = single selects, reads = range tests, xacts = transaction tests, etc.  Transaction tests are counting individual queries, not transactions.  The "Rdm" tests are using a uniform distribution, whereas the non-'Rdm' tests are 75% of queries are using 10% of the rows. Â
Here are the results:
It's always hard, of course, to see what these numbers would look like on a real production database. Â The numbers for the select queries look pretty decent, but a real database probably wouldn't just be doing those all day. Â It seems clear that a lot of writes into the db will limit the I/O scalability pretty quickly, which shouldn't really be a big surprise. Â
UPDATE:Â Â I'm pretty sure these results are skewed by sysbench having it's own set of bottlenecks.
Trackback URL for this post:
http://mysqlguy.net/trackback/23
One question, which always
One question, which always warrants more benchmarking is how Linux co-oeprates with the external RAID device with respect to scheduling. I'd like to see numbers based on the different types of schedulers. Also, what filesystem was used in this test?
Yeah, I've found good
Yeah, I've found good benchmarking tends to lead to more benchmarking. (Just like a good Science experiement answers some questions, but reveals more).
I'll double check, but I believe this test is using the deadline scheduler. The filesystem is ext3. The MSA-30 in this test is plugged into the second channel of the same RAID card doing the internal RAID... I wouldn't expect it to perform much differently as a result of that. Oh, and battery-backed write cache is enabled.
Post new comment