ADSM-L

Re: [ADSM-L] FW: Somewhat OT: Sizing a VTL solution

2007-06-21 17:55:00
Subject: Re: [ADSM-L] FW: Somewhat OT: Sizing a VTL solution
From: "Johnson, Milton" <milton.johnson AT CITI DOT COM>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Thu, 21 Jun 2007 17:53:35 -0400
Yes that is an approach that will work. I used:
(Total_Bytes_in_Primary_STG + ((Max_Daily_Amt_Backed-up X (ReuseDelay +
2))) X Growth_factor

I then let the compression factor be my fudge factor.

However give some thought to the size (native capacity) of your virtual
tape volumes.  In the physical world you have little control over this
(we generally use the largest size cartridge we can get to have the
greatest capacity [volume size times fixed number of cartridge slots =
total capacity online]).  In the VTL world things are different, you
total capacity is fixed (determined by the disk space you purchased) but
YOU can determine the volume size and the number cartridge slots equals
the total capacity divided by the size of your virtual tape volumes.

This said, what can you do with it?  I went with a small tape size
(10GB) and have the following benefits:
A) If I need a file at the end of the tape I spin through 10GB not 100GB
of data (ignoring compression).
B) No matter what reclamation threshold you use, you will have tapes
that are 1% short of being reclaimed.  Small tapes in this state "waste"
less space then large tapes in that state.

One gotcha is deciding which physical library you are emulating.  When I
first brought my VTL up, TSM choked.  Why?  Because TSM "knew" that the
library I was emulating could not have 1,415 physical tape slots.  I
changed the emulation to an "ADIC Scalar 10K" and the problem was
solved.  Sorry I don't remember the original library emulation.

Just a thought, think outside the physical box, it's all virtual.

Thanks,
H. Milton Johnson
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
wanda.prather
Sent: Thursday, June 21, 2007 2:35 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: [ADSM-L] FW: Somewhat OT: Sizing a VTL solution

I agree!  That's the method I would use.

TSM thinks the VTL is a tape library, and processes it as such.
So you need to look at how many volumes you are using, not just how much
space.

If the numbers you see for volume usage don't square with your
occupancy, better find out why!
In most customers, I find their overall tape utilization is at best
around 65% full (because there are tapes in FILLING status, DB backup
tapes, EXPORT tapes, scratch tapes, tapes that are below the reclaim
threshold, etc.)

And as far as I can tell, if you set your reclaim threshold to 50%, over
time you can, quite by accident, end up with hundreds of tapes sitting
at 49% that never reclaim without manual intervention.  You can be more
aggressive about running reclaims in a VTL.  But you still have to
account for un-reclaimed tapes and volumes that never fill (like DB
backups).

Now I know some VTL's will allocate their physical storage in chunks
(say 5G for example) as a tape is written/appended to.  So if your
virtual volumes are 100G, but your DB backup only needs 20G, you will
only be using 20G of the backstore.  But if your DB backup needs 21G,
you will be using 25G of the backstore.  SO it is possible to get away
with creating more virtual volumes than you actually have space to
support.  But I'm very leery of attempting to overcommit volumes in the
VTL, given that TSM tries its best to create full volumes.

What I would like to know is if anyone has good ideas on what level of
overcomittment you can get away with?  Or is it better just not to go
there?

W


  _____

From: ADSM: Dist Stor Manager on behalf of Andy Huebner
Sent: Thu 6/21/2007 2:21 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: Somewhat OT: Sizing a VTL solution



The approach I used worked very well.  I counted all of the tapes that
would be kept on the VTL and multiplied that by the capacity of the
tapes.  I then calculated the growth rate of the tape pool and added
that to the total, in my case 6 months out.  That is the number I used.
The assumption is that compression will be similar on both.

If you calculate off of occupancy there is a greater uncertainty because
you do not really know how much compression you will get.  Calculating
off of tape counts will be inaccurate because of filling tapes, but this
error is in the admins favor.  Filling tapes can be factored in to
reduce the error.

Andy Huebner

-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
William Boyer
Sent: Thursday, June 21, 2007 1:03 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: [ADSM-L] Somewhat OT: Sizing a VTL solution

Has anyone just gone through sizing a VTL solution for a library
replacement? Is it as simple as taking your current occupancy/retention,
applying for some compression and using that figure for the amount of
storage behind the VTL? Or maybe I'm just trying to make something
harder than it is. Wouldn't be the first time!
:-)

TIA.

Bill Boyer
>Select * from USERS where CLUE>0
0 rows returned


This e-mail (including any attachments) is confidential and may be
legally privileged. If you are not an intended recipient or an
authorized representative of an intended recipient, you are prohibited
from using, copying or distributing the information in this e-mail or
its attachments.
If you have received this e-mail in error, please notify the sender
immediately by return e-mail and delete all copies of this message and
any attachments.
Thank you.

<Prev in Thread] Current Thread [Next in Thread>