HP3000-L Archives

May 2004, Week 2

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
John MacLerran <[log in to unmask]>
Reply To:
John MacLerran <[log in to unmask]>
Date:
Thu, 13 May 2004 14:00:37 -0600
Content-Type:
text/plain
Parts/Attachments:
text/plain (92 lines)
Hi Denys,

I looked at what you had posted earlier, and attempted to configure the
array that way, but it wouldn't let me. After the 8th LUN, I got a
message that said "Invalid LUN size" and it wouldn't let me go any
further (I tried increasingly smaller chunks until I was able to create
a tiny 400mb LUN). We want to use RAID 1/0 topology, which is basically
disk mirroring, so we effectively get only half the total amount.  Ten
36gb discs is 360gb, of course, so what I'm seeing when it runs out of
space at around 165gb is probably a factor of the 165gb being mirrored
(330gb) plus some overhead space for cache mirroring or other system
level voodoo. Thus the question of how best to carve up that 165gb to
get the best performance.


Denys Beauchemin wrote:

>I posted the following fragment in answer to a similar question back in
>early April:
>
>"If the controller will be connected to an FC card on the 3000, you
>should create a maximum of 16 LUNs (ldevs) (20GB each).  If the array
>will be connected via an HVD or LVD SCSI connection, plan on a maximum
>of 8 LUNs (ldevs) (40GB each).  If it's connected via SE-SCSI, plan on
>updating your resume.  Just kidding, don't do it that way."
>
>Obviously, you need to recalculate the LUN sizes given your array size.
>
>
>Denys
>
>-----Original Message-----
>From: HP-3000 Systems Discussion [mailto:[log in to unmask]] On
>Behalf Of John MacLerran
>Sent: Thursday, May 13, 2004 12:34 PM
>To: [log in to unmask]
>Subject: [HP3000-L] VA7100 Questions
>
>Dear HP-3000-L,
>
>We just received a refurb VA7100 Array to replace two Nike Mod 20s on
>our
>N4000 MPE box (woo hoo!). We purchased it with ten 36-gb discs, and I
>have a
>couple of questions about how best to partition the lil' beastie for
>performance.
>
>I've read that MPE gets the best performance wrt disc i/o by having
>"lots of
>small luns" -- my questions are:  How many is "lots"? and How small is
>"small"?
>
>By trial and error, I've discovered that the current raw configuration
>of
>the array yields about 165GB (in RAID 1/0, which is how we'll use it,
>because we  don't want the RAID5 write performance hit). I could
>partition
>that into fifteen 11gb luns, or eleven 15gb luns, or twenty 8gb luns and
>a
>5gb lun. If I want to go really crazy, I could do fifty-five 3gb luns,
>but
>that seems overkill. Is there a point of diminishing returns?
>
>Also, given that MPE 7.5 can use space beyond 4gb on LDEV1, is there an
>advantage to keeping that lun smaller than the others (the 5gb lun in
>the
>20/8 scenario above)?  Assuming, of course, that I can tell the array
>which
>lun I want to correspond to LDEV1 when I run mapper and fcscan after we
>get
>the adapter cards installed.
>
>Any insight?
>Thanks!
>
>* To join/leave the list, search archives, change list settings, *
>* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
>
>
>

--
----------------------------------------------------------------------
  John MacLerran
  IT Systems Analyst                       email:   [log in to unmask]
  Idaho State University                             V(208) 282-2954
  http://www.isu.edu/~macljohn                       F(208) 282-3673
----------------------------------------------------------------------

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

ATOM RSS1 RSS2