System : Centos 6.4
, PerconaDb/MySQL 5.5.31
Im running relatively small mysql database with some 30 databases, total amount of 12Gb, but heavy on reads, updates and writes (texts, posts, user data and user actions are being CRUD evenly during the day). Its a typical relational db, no data warehouse or so.
Currently, its running on some two HDD array (not sure about specs, its hosted) giving the disks quite tough time, with 25% IO utilization, avg 700 IOPS, 5Mbps (according to Newrelic monitoring agent on the server). The whole MySQL
server is doing some 540 queries per second, average.
Im considering putting it on separate disk array, and Im considering use SSD for it. Either classic SSD disks, or SSD PCI card.
As there are technical limits of how much data can be written to SSD, before the chips wear out, I'd like to calculate / estimate the amount of time the disks will be good before need to be replaced.
My questions are :
- Is it premature optimization ? I cannot rewrite all apps using all the databases, all are indexed as well as possible, I'm no newbie in this, but I accept there might be things to consider and change.
- How can I measure the amount of data written by
MySQL
? to estimate how long it will probably live on the SSD. Does it even apply on this amount of data, should I care about it at all ? - Do you have any experience with running
MySQL
on SSD, any best practice to share in here ?