Rob,
A unload/reload would be too long of a down window..
Also, I don't have times but the below could be another 5 to 10 hours.
Anyone have times for the various SSD drives, 177, 387, 775?
Start Raid
Init disks
Add to ASP
Paul
-----Original Message-----
From: MIDRANGE-L [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of rob@xxxxxxxxx
Sent: Monday, December 22, 2014 10:27 AM
To: Midrange Systems Technical Discussion
Subject: RE: Will mixing 775 SSD and 177 SSD in a 5887 EXP24S I/O drawer be a potential performnace issue
I've done it without an unload/reload. And I've done it with an unload/reload. Frankly I think the unload/reload is easier, less expensive and has less IPL's. And what IPLs it has is all at once instead of "run this for awhile, IPL, run this, IPL... To me it's easier to coordinate with getting it all done one weekend while Mimix swapped to backup machine versus planning several midweek IPL's.
Rob Berendt
--
IBM Certified System Administrator - IBM i 6.1 Group Dekko Dept 1600 Mail to: 2505 Dekko Drive
Garrett, IN 46738
Ship to: Dock 108
6928N 400E
Kendallville, IN 46755
http://www.dekko.com
From: "Steinmetz, Paul" <PSteinmetz@xxxxxxxxxx>
To: "'midrange-l@xxxxxxxxxxxx'" <midrange-l@xxxxxxxxxxxx>
Date: 12/22/2014 09:59 AM
Subject: RE: Will mixing 775 SSD and 177 SSD in a 5887 EXP24S I/O
drawer be a potential performnace issue
Sent by: "MIDRANGE-L" <midrange-l-bounces@xxxxxxxxxxxx>
Sue,
I never thought of the odd number of parity set issue.
I currently have 2 sets of Raid-5, my plan would result in 3 sets.
I think I should re-evaluate, come up with an option which includes an
even number of parity sets, same size drives.
Using either all 387 SSD or all 775 SSD.
Is there a larger SSD on the horizon?
I did a similar process back in 07 on a P5, went from 35 to 141.
I did this without a unload/reload, migrated the load source and moved all
the data to the new larger drives, minimize the down window time.
I was planning on repeating this same process.
I have a 2nd EXP24S that is filled with 24 571 10K spinny, used for an R&D
LPAR.
I was thinking of a creative process where I could temporary unload this
IO drawer to load up the new drives, then swap everything back when
complete, without doing a unload/reload.
Paul
-----Original Message-----
From: MIDRANGE-L [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Sue
Baker
Sent: Monday, December 22, 2014 8:11 AM
To: midrange-l@xxxxxxxxxxxx
Subject: Re: Will mixing 775 SSD and 177 SSD in a 5887 EXP24S I/O drawer
be a potential performnace issue
"Steinmetz, Paul" <PSteinmetz@xxxxxxxxxx> wrote on Mon, 22 Dec
2014 03:39:18 GMT:
On a P7, 740 8205 E6C, I currently have 18 177 SSF-2 SSD (2844 gb;
5913 controllers) in a 5887 EXP24S I/O drawer, which is 100% of my
production LPAR, CEC is not used for this LPAR. We need to add disk
for a few new PF and LF that will take a large amount of DASD. I have
6 empty disk slots in a 5887 EXP24S I/O drawer. I was planning on
adding 6 775 SSF-2 SSD
(3875 gb), one new RAID-5 parity set.
How manhy parity sets do you currently have? If creating 1 new parity set
puts you in the position of having an odd number of parity sets, this
could cause an imbalance for the controllers which could create
performance anomalies.
1) I could add these 6 drives to the current sys ASP.
These 6 drive would be accessed much more heavily because of
the large difference in size. Could this be a potential
performance issue, because these would be "hot drives"? "Hot
drives" with SSD would not be as noticeable as with spinny hot
drives.
The new SSDs have better performance than the old 177GB SSDs.
This may be enough to offset the effect of mixing the two SSD
sizes.
Or,
2) I could create a new ASP for these 6 drives, and then
place only these new files in this ASP. Still have the hot
drive issue, but less impact.
But now you have to manage files in different ASPs. This can
lead to complications for DB management, journaling, etc.
3) Are there other options??
midrange.com
