No, SFS will not do concurrent I/O to minidisks located on a single volume.
SFS knows on which volules its minidisks are located and acts accordingly.
If I remember well, I displayed some minidisk usage in my first append for
this subject, where one can see that SFS leaves mindisks unused. Have a
look at the description of y SFSULIST package:
http://www.vm.ibm.com/download/packages/descript.cgi?SFSULIST
I discuss the subject a little bit and show an example of minidisk usage,
read the section from "the Q FILEPOOL MINIDISK command" on.
2007/6/21, O'Brien, Dennis L <Dennis.L.O'[EMAIL PROTECTED]>:
Steve,
I know an SFS server won't do multiple I/O's to the same minidisk. My
question is whether it will do concurrent I/O's to different minidisks on
the same volume. If it will, there might be some benefit to putting two or
three minidisks on one volume and using PAV. If not, I'll just put one
minidisk on each volume, or put minidisks from different SFS servers on a
volume.
Dennis O'Brien
"I don't have a girlfriend. I just know a girl who would get really mad
if she heard me say that". -- Mitch Hedberg
------------------------------
*From:* The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] *On
Behalf Of *Steve Wilkins
*Sent:* Thursday, June 21, 2007 06:58
*To:* IBMVM@LISTSERV.UARK.EDU
*Subject:* Re: [IBMVM] PAV and SFS
PAVs will not benefit SFS. An SFS server (1 entity) can't be setup to do
multiple I/Os to the same minidisk. Therefore, the VM I/O dispatcher doesn't
have anything (ie. multiple I/Os) to spread out over the real system
attached Aliases.
Regards, Steve.
Steve Wilkins
IBM z/VM Development
[image: Inactive hide details for Marcy Cortes
<[EMAIL PROTECTED]>]Marcy Cortes <
[EMAIL PROTECTED]>
*Marcy Cortes <[EMAIL PROTECTED]>*
Sent by: The IBM z/VM Operating System <
IBMVM@LISTSERV.UARK.EDU>
06/21/2007 12:10 AM Please respond to
The IBM z/VM Operating System <
IBMVM@LISTSERV.UARK.EDU>
To
IBMVM@LISTSERV.UARK.EDU
cc
Subject
Re: PAV and SFS
It depends!
I think you have to study your perf data and decide where your i/o
contention is. We don't have it like we used to (pre IBM DS8000 and
FICON
) --- have moved on to just CPU and memory issues :)
We've moved most of our disks to mod 9 (all of prod Linux and about 1/2
the
traditional CMS users) and the last 3 TB that went in I had them do mod 27
(makes creating large linux filesystems easier). Not using PAV anywhere.
We just don't see the i/o queue that would warrant the effort.
On the dev/test system with 100 linux virtual servers and not enough
memory
(24G real), we have about 50 mod 3's for paging. I don't suspect moving
that to 16 mod 9's would be that much different. 16 is still pretty
spread
out. Now, if I had 3 page vols, I woudn't dream of putting them on 1 mod
9
:)
It is a bit confusing for the perf guys though ---- since disk size does
not
show up in the perf data!!!
And, I understood SFS to work like Kris said. We have both 3s and 9s in
our
SFS - we always give it the whole vol (minus cyl 0 of course ).
Marcy Cortes
"This message may contain confidential and/or privileged information. If
you
are not the addressee or authorized to receive this for the addressee, you
must not use, copy, disclose, or take any action based on this message or
any information herein. If you have received this message in error, please
advise the sender immediately by reply e-mail and delete this message.
Thank
you for your cooperation."
-----Original Message-----
From: The IBM z/VM Operating System
[mailto:IBMVM@LISTSERV.UARK.EDU<IBMVM@LISTSERV.UARK.EDU>]
On
Behalf Of O'Brien, Dennis L
Sent: Wednesday, June 20, 2007 6:32 PM
To: IBMVM@LISTSERV.UARK.EDU
Subject: [IBMVM] PAV and SFS
Our DASD people have asked if they can replace our 3390-3 size DASD
volumes
with 3390-9's. My tentative answer is that most of the volumes that
contain
minidisks can be moved, if the new DASD has PAV. CP areas such as paging
should remain on 3390-3. Our SFS servers have 3338-cylinder minidisks,
one
per volume. My initial thought is that I could put 3 of them on a 3390-9
with two PAV aliases, and the SFS server could do one I/O to each minidisk
at the same time, just like it can when the minidisks are on separate
volumes.
In the recent "dasd 3390 -27" thread, Kris Buelens said, "For SFS though
(and probably DB2) PAV is maybe of limited help: a given SFS server will
start only one I/O to all its minidisks on the same disk." Is that really
true? Does an SFS server look at the real devices or volsers that its
minidisks are on, and only start an I/O to one minidisk on a volume at a
time?
Dennis O'Brien
The three R's of Windows problem resolution: Retry, Reboot, Reinstall.
--
Kris Buelens,
IBM Belgium, VM customer support
