RE: I/O tuning... Allocating spindles to databases

>>>
>>>In retrospect, you're right, I could see how the presence of
>>>cache could degrade performance (a little).

I have done TPC-D on SAN systems moving 3.5 GB/s ...AFTER disabling array cache... and, uh, let's say significantly less with it turned on.

It has historically been a problem. To deal with it, the high end array manufs started trumping "adaptive cache" which often times meant a bypass read.

I'm not saying array cache is bad. I'm just saying don't presume it is good.

I have a pretty large joint IBM whitepaper coming out where we sis a lot of this sort of testing with the TotalStorage DS4X00 stuff... it is always interesting to play

--
http://www.freelists.org/webpage/oracle-l