On 12/1/2010 7:54 AM, Kevin Grittner wrote:
I agree that storage makes a bigger difference than filesystems in most cases. Just wanted to jump in and say that testing with your workload is really the only way to know things for sure. (Obviously, more spindles implies more IOps, more cache implies faster writes...) But I have seen cases where a dell PERC RAID controller (not sure that qualifies as "good") slowed things down when configured for writeback with a write heavy workload. We noticed very spiky disk IO rates on some database servers (we are monitoring company, so have lots and lots of pretty graphs.) The controller seemed to accept a bunch of writes quickly, then when full then try to pump them all to disk, which took a non-trivial number of seconds. So we'd see the interesting state of iostat reporting zero disk transactions per second, but busy time = 100% and service time being hundreds of millseconds. Switching to write through made things much faster overall. Not what I expected. This was a PERC 5, so a few years ago. No doubt things have improved, but the point is to test and monitor. --
|