I have a chance to test a system with Intel 320 SSD drives (NewRelic provided me with an access to the server), and compare performance with SAS hard drives.
For the benchmark I took a sysbench uniform oltp rw workload. 256 tables, 50mil rows each, which gives in total 3T of data.
To vary a ratio memory/data I will vary an amount of tables from 256 (3TB) to 32 (375GB).
As a backend database I use Percona Server 5.5.19.
I should mention that on these datasizes, sysbench workload is pretty nasty, MySQL will mostly reads and writes pages from buffer pool (replacing pages in buffer pool). This however allows us to see the best possible scenario for SSD running under MySQL, the final result will show the best possible gain.
I do measurements every 10 sec to see stability of results.
Tables HDD SDD Ratio
32 1226 1644 1.340946
64 140 571 4.078571
96 101 506 5.009901
128 89 486 5.460674
192 79 484 6.126582
256 75 495 6.600000
As you can see, on the big datasizes we have 5-6x improvement. However on 32 tables (375GB of data), the result became unstable.
It looks like we are having symptoms of the flushing problem. This is to investigate later.
The scripts and raw results are on Benchmarks Launchpad.