Estimating IOPS
From: NetComrade (netcomrade_at_bookexchange.net)
Date: 12/22/04
- Previous message: Levi Ashcol: "Process Sorting Criteria !"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Wed, 22 Dec 2004 15:27:49 -0500 (EST) To: sunmanagers@sunmanagers.org
Reposting from VxVM group.. Let's see how many 'I am on vacation' replies
I get.
As you might've seen from my previous posts, we've been trying to get
A5200's to work w/ Linux. (we have 4 in production, having Oracle running
on top, each array basically belongs to a server, they're connected to
hubs, and can failover from server to server)
Looks like at this point, we'll be looking to replace arrays with
something else. Since we must shop on a very tight budget, we will most
likely have to reduce the # of spindles. In order to do that, we'll need
to figure out current throughput and IOPS of current production usage.
The most simple way I thought of was to look at
a) iostat
b) vxstat
Vxstat seems to make much more sense, since we'll be getting IOPs per
volume, which I could store in the database and query for later. Once the
'max' is determined (let's say over a month), it'll be probably 2x-4x to
determine 'minimum sustained IO operations per second' for a new disk
array. The logistics of it are not so straight-forward, I don't think
vxstat is capable of acting like sar and storing this stuff somewhere, or
is it?
Has anyone done an assessement of their current IO needs? Anything else I
should be looking at? Any (free) tools would do this for me?
Thanks.
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
- Previous message: Levi Ashcol: "Process Sorting Criteria !"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|