All
I should have put more thought into this rather than just copying some
text with the formula.
John made me realize that I should have added my personal thoughts about
this. Personally, I know of no one with the patience to calculate an
alias value for each and every LCU in their shop. They might be
different and God knows that the data might move around across LCUs
during the course of a week and the SSCH value might be different today
than it was yesterday. SOOOOO, having said that, I usually recommend
taking the highest SSCH value for an LCU and using that value as a
constant for all LCUs. This assigns the largest expected alias count
for an LCU to all LCUs. If the data moves around, you are covered.
Also, this formula typically results in much lower values for alias
definitions than already exist, if you calculate a value that is too
high, it doesn't hurt -- you are still using less than before. I mean
think about it -- my previous recommendation for customers with mod-27s
in an LCU was 3 alias to 1 base, so you would allocate 64 base devices
and 192 alias devices -- with HyperPAV you can possible reduce the 192
down to 20. What a bargain even at twice the price.
Tom
John Ticic IBM-MAIN wrote:
You are moving to a new box, so you need to look at your current peak I/O load,
for each LPAR and calculate (on an LCU basis) how loaded your new LCUs will be.
The method shown below is a good starting point, but pick your factor to cope
with bursts of I/O activity. Since HyperPAVs are assigned on an I/O basis, you
should look for the peak I/O workload (not the sum of all LPARs!).
You should be conservative and use the RMF records (74.1 and 78.3) to verify
your calculations (DASD device and I/O queing reports) after you have moved to
the new box.
Take a look at DS8000 HyperPAV case study (IBM techdocs) and also on our
website for a white paper that will give you some background information.
John
2 * [ (SSCH Rate) * (Average I/O Service Time) ] = # of Proposed Aliases
Example -- assumptions are 2000 SSCH for the LCU with an average I/O
service time of 5 milliseconds.
2 * [ (2000) * (.005) ] = 2 * 10 = 20 Aliases for the LCU
Note ... Make this calculation at the LCU level as the alias are defined
as device numbers on an LCU.
Hello,
We are installing a new DASD array with 33 TB of capacity with
approximately 1/3 of the capacity being setup as 3390-3's, 1/3 as 3390-9's
and the last 1/3 as 3390-27's. I am wondering if anyone has any experience
or documentation that would suggest how many Hyper PAVs per LCU we should
define. The LCUs will contain an even mix (from a capacity perspective) of
the 3390-3's, 3390-9's and 3390-27's.
John Ticic
IntelliMagic - Storage Intelligence
Perzikweg 13a, 2321 DG Leiden, The Netherlands
www.intellimagic.net
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html