Glenn,

I work for one of the three vendors, and so I am very surprised to hear that 
HDS thinks that "Tape is clearly the best storage media for long term data 
archiving." Would you happen to know who it is that represents HDS 
hardware/software that agrees with this? I for one strongly disagree with the 
statement.

Why would you suggest that CU tiering is only for small environments? I'm 
thinking the large, multi PiB environments will gain far more from storage 
tiering than smaller to medium shops. They have far more to gain from removing 
the overheads of Primary and Secondary space management and the costs of 
recalls, especially when they can take advantage of commodity midrange storage 
for inactive data. This is the same storage that makes virtual tape so 
attractive, but without the cost of redundant data movement and transformation. 
The two single greatest changes in DFSMShsm in recent times have been CRQ and 
Transitioning, where CRQ acknowledges the problems of managing large scale HSM 
activity, and transitioning is a kludge that allows DFSMShsm to take advantage 
of CU based tiering. In fact I think I spoke to you in San Francisco about the 
possibility of replacing the transitioning command with a proprietary migration 
command if the vendor had a way to do command level tiering of a data set 
extents rather than FlashCopy.

The DFSMShsm ML2 strategy decrees that data can no longer be directly accessed 
when it becomes inactive or dormant. Why is that a better strategy for 
Information Lifecycle Management than allowing the data to be accessed directly 
even as it moves to a more cost effective media as it ages? Why is last 
accessed a date a better strategy for archiving data than the backend IO/hour? 
Why can't I archive 100s of TBs of data because someone keeps opening the file 
but not touching the data? I agree that "Each tiering technique, hardware and 
software, has strengths and weaknesses" but I earnestly feel that the archiving 
controls afforded DFSMShsm do not lend themselves to good storage cost 
containment. I acknowledge again that ML1 and disk based ML2 can operate very 
effectively in a tiered storage environment, especially where ML2 can 
transparently migrate to commodity midrange storage as it ages. I do not 
however agree that the cost of data translation and recall are necessary to 
maintain good ILM strategies.

I still remember the Iceberg pundits from STK describing how there would be a 
Nearline library behind every Iceberg and HSM would disappear. I think we are 
getting closer to this every day and many shops are in a position to do exactly 
that.

Ron


-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf 
Of Glenn Wilcock
Sent: Tuesday, August 26, 2014 10:05 AM
To: [email protected]
Subject: [Bulk] Re: [IBM-MAIN] [Bulk] Re: [IBM-MAIN] General question on moving 
DFHSM work from mix TAPE/DASD to More DASD

While control unit storage tiering may be considered as a replacement to HSM 
processing for smaller environments, such a recommendation is an over 
simplification of the need for a comprehensive ILM strategy to properly manage 
data in middle-to-large environments.  At the various conferences that I attend 
each year, this concept was originally discussed when cu tiering was first 
introduced, but after discussions, all three vendors see the value of HSM ILM 
and cu tiering being used together to create a powerful solution as opposed 
trying to select one over another.  Each tiering technique, hardware and 
software, has strengths and weaknesses.  Using each technique to its strengths 
provides tremendous opportunity as we move forward with managing the 
significant growth of data that we are seeing.  In z/OS V2R1, DFSMS introduced 
its initial Storage Tiering solution.  This offering lays the framework for 
z/OS's long term strategy to provide various ILMs solutions so that clients can 
implement the ILM solution that works best for them.  An integral part of this 
strategy is to move away from ML1 and move toward an L0 - Ln, ML2 solution.  
Tape is still clearly the best storage media for long-term data archiving, and 
all three vendors will agree to that.  I am currently working with clients to 
move to an L0 - Ln, ML2 environment, and it is exciting to see the 
opportunities that exist by integrating software and hardware tiering into a 
single, powerful ILM strategy.  I'm more than happy to meet with clients to 
discuss the V2R1 DFSMS Storage Tiering solution and discuss the opportunities 
that it provides to exploit the strengths of the two types of tiering.

Glenn Wilcock
DFSMShsm Architect 

----------------------------------------------------------------------
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