Since all writes and sequential reads are cached by the EMC storage system, the best use for solid state disk is to store data that has a random READ pattern. On Oracle, that is the temp tablespace which is used for disk sorts that cannot fit in memory. How can I isolate the parts of MySQL which are characterized by a random disk read pattern?

Here's a clarification of my original question: This is a reporting database with very large tables, large sorts, etc. Cannot fit into server memory.

  • You can buy off the shelf now boxes with 64G or 128G RAM in them - I have a a hard time believing that it makes economic sense to use SSD for sorting.
    – Gaius
    Jan 12, 2011 at 22:10

1 Answer 1


Typically any OLTP database that can't fit into memory has random reads. If you are reading that much from your OLTP database that you need EFDs (what EMC calls SSDs) then you need to first look at putting a LOT more RAM into the database server. It will be a lot cheaper to prevent the reads than to make the reads faster. Those EMC EFDs run about $20k each.

  • Thanks for your answer. I should have been more specific: large data warehouse, huge tables, lots of sorting. Jan 12, 2011 at 16:16
  • I updated the original question to make it more specific. Jan 12, 2011 at 18:28
  • 1
    Actually all reads are cache by the EMC array, unless you disable the read cache for that LUN (easy to do, just a check box). With a large high IO database EFDs can definitely reduce your IO wait times. I'll leave the MySQL stuff to someone else (I know almost nothing about MySQL at this point).
    – mrdenny
    Jan 12, 2011 at 20:07
  • All reads are cached but the cache will not have the next random read and the one just cached won't be read again. We are getting a new Storage System with EMC FAST so all the hot blocks will migrate to SSD. So the question is moot now. Jan 21, 2011 at 21:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.