Lizette,

With Tiered Storage on internal disk available on all three vendors, and on
midrange storage with HDS, perhaps it is time to re-examine the need for ML1
with the big 3, and ML2 with HDS.

The objective of ML1 is to compress recently inactive data sets and use less
GB on media that is the same $/GB as Primary storage. The objective of ML2
is to move long term dormant data sets to a cheaper media. Both processes
require data movement and transformation to archive and recall the data sets
which introduces overhead and recall latency.

Don't shoot me for using the HDS example - it's who I work for and what I
know best. A three tier strategy using HDD or SSD for Tier 1, Nearline SAS
for ML2, and virtualized Brand-X midrange storage for Tier 3 presents a new
paradigm for archiving inactive and dormant data sets, including back-ups,
that I believe over time can displace DFSMShsm altogether.

It is the cost per GB of the media that provides the savings, divorced from
double handling and transformation of the data. Data sets are written to the
first tier of a tiered storage pool much in the same way as they are written
to Primary storage with DFSMShsm - there are no real changes here. What does
change is that from this point on the storage is managing pages, not data
sets, and as pages go inactive they will move to cheaper NL-SAS installed in
the same controller, and finally onto virtualized Brand-X midrange storage -
perhaps a repurposed IBM-XIV from and Open application that the bean
counters want kept for a another year, or something you got cheap with a
bunch of squatty boxes from Oracle. 

What I find important is there is no data transformation or recall latency:
it is all transparent to the application. You have to read 12 months of
General Ledger files or SMF data sets? The application simply reads them
directly from whatever tier disk the pages happen to be on. You're not
waiting 24 hours for data sets scattered all over myriad ML2 tapes to be
recalled, you don't have to find redundant Primary space to store the
recalled data sets, and there won't be any TMM thrash when they are migrated
again by the next space management cycle. That process is transparent to
z/OS in the way we thought the STK Iceberg would go. Of course all this
dormant data remains replicated with TC and/or HUR while it is being
shuffled around the backstore tiers in both sites - you're not moving any
DFSMShsm traffic across the replication links.

The big saving that I see here is vitalizing the cheapest, acceptable
midrange disk for Tier 3. It looks and smells like a 3390 to System z, but
it is just commodity midrange disk that you buy at the local midrange disk
vendor store. You're no longer locked into the big 3, but you can still use
it with TrueCopy, HUR, Shadowimage, FlashCopy, etc through the Enterprise
Controller where the backstore is a HP 3PAR storage system. Whatever gives
you the best bang for your buck.

This probably sounds like a bit of a hard sell, but I drank the cool-aid
back in 1994 with Iceberg and I think DFSMShsm is looking rather old and
heavy compared to Storage tiering from IBM, EMC and HDS.

Ron

-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[email protected]] On
Behalf Of Lizette Koehler
Sent: Saturday, August 23, 2014 12:19 PM
To: [email protected]
Subject: [Bulk] Re: [IBM-MAIN] General question on moving DFHSM work from
mix TAPE/DASD to More DASD

The driving force is the Mgt team is familiar with NetBackup and Other Open
system software that does backup.  They do not understand mainframe.

I have been unsuccessful in explaining it is MORE THAN BACKUP on the
mainframe.  And we cannot lose HSM Data unlike NetBackup.

So that is the process driving the discussion.  I am concerned I cannot
anticipate how much spinning disk they would need on the storage array to
replace a tape appliance for tape datasets.  And HSM is our biggest
consumer.


Lizette


> -----Original Message-----
> From: IBM Mainframe Discussion List [mailto:[email protected]] 
> On Behalf Of TonyIcloud-OPERA
> Sent: Saturday, August 23, 2014 9:04 AM
> To: [email protected]
> Subject: Re: General question on moving DFHSM work from mix TAPE/DASD 
> to More DASD
> 
> Yes, 1988. We had abundant spinning DASD since that's the only kind 
> that
existed.
> It was a small MVS shop that IBM provided ($$) generously.
> 
> Usual questions:
> 1. Do you have enough disk?  No guessing allowed.
> 2. Is the tape mount scenario driving this plan?
> 2. Will it complicate your DR?
> 3. Will it complicate your routine non-DR backups?
> 
> I know you asked that below but every shop has its own methods of 
> backup
for DR
> and non DR.  Without intimate knowledge of your current backup 
> methods.......speculation.
> 
> All in all I think it's an excellent idea, subject to niggling details 
> of
course.
> 
> 
> 
> On Fri, 22 Aug 2014 22:42:38 -0500, Lizette Koehler
<[email protected]>
> wrote:
> 
> > We are z/OS V1.12
> >
> > We are considering changing from a mix of Tape and Disk for HSM 
> > functions.
> >
> > So ML2 would go away or be very reduced
> > ML1 would be expanded
> > BACKUP and ABARS would be directed to DASD.
> >
> > Has anyone done this?
> >
> > If so, what were some of the experiences?  Did you find the increase 
> > in dasd requirements were exponential?  Or were they an easy slope 
> > (that is not any quick growth) Did you find any issues with 
> > constantly adding dasd?  Or could you stabilize the HSM needs.
> > Did it provide better support for Business Recovery or Disaster
Recovery?
> > Any performance issues?
> >
> > Thanks for any input
> >
> > Lizette
> >
> > --------------------------------------------------------------------
> > -- For IBM-MAIN subscribe / signoff / archive access instructions, 
> > send email to [email protected] with the message: INFO 
> > IBM-MAIN
> 
> 
> --
> Using Opera's mail client: http://www.opera.com/mail/
> 
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions, send 
> email
to
> [email protected] with the message: INFO IBM-MAIN

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions, send email
to [email protected] with the message: INFO IBM-MAIN

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to