I saw a few articles ([url]http://www.percona.com/files/white-papers/fusion-io-mysql-multi-instances-report.pdf[/url] , [url]http://www.percona.com/files/white-papers/scaling-multiple-mysql-percona-server-virident.pdf[/url]) which compares single vs multiple MySQL instances performance. But all of them speak about cutting edge storage like FusionIO and so on.
Are there similar tests for the old SAS disks?
TIA,
Vitaly
In my opinion it makes little sense to practice this on server running spinning disks (whether those are 7.2krpm sata or 15krpm sas), basically because usually it’s the storage the biggest bottleneck for most workloads. I could think of dealing with concurrency issues for in-memory workload this way, where you could split read traffic between two instances on the same box. But this could bring some improvements only in some specific cases, like small write traffic, all data fits in (x times smaller) buffer pool, lot’s of CPU cores available, etc.
The reason why it is sometimes worth to run multiple MySQL instances on the same machine with Fusion-IO, Virident, and other very fast PCI-e flash cards, is that single MySQL instance is hardly able to utilize total available I/O throughput in many workloads.