From: "Bob Elzer" <bob.elzer@gmail.com>
> What about a large database, maybe a couple of terabytes, and maybe
> lots of users. And then let's say we spread our 6G of swap over
> several disks. Maybe a system with 32GB of memory, and quad
> processors. Maybe we could run some tests, anybody have a system
> like this?
Lessee. Not quite, but: Master DB server over here is 4-CPU, 16G
RAM, has ~90G of junk in InnoDB and MyISAM tables, has hardRAID-5,
and swap usage is currently roughly constant at 20M. We aren't
Amazon or NewEgg, but we get a constant flow of insert/update/delete
statements. 200-300 queries/second, mostly. Similar stats on the
read-only slaves, which have practically identical hardware specs.
When things are working like they should be, swap usage is minimal.
I've commented out the swap lines in fstab on a few machines, and
seen no problems in those swapless machines. Occasionally,
I've seen the Apache boxes hit swap heavily, fall over, and barf,
but that's been a lot less common since we changed a few things
early this year. YMMV, obviously.
--
Matt G / Dances With Crows
The Crow202 Blog:
http://crow202.org/wordpress/
There is no Darkness in Eternity/But only Light too dim for us to see
---------------------------------------------------