One of the recommendations for SMS used to be to use the allocation
space as a factor in mapping a STORCLAS to different storage groups
based on size in order to reduce impact of fragmentation and reduce the
need for volume defrags. Using this technique to direct larger files to
mod 9's or mod 27's can keep the file count on these devices modest and
reduce need for humongous VTOC/VTOCIX, since you no longer have the
possibility of 100,000 tiny files. We actually ended up with larger
VTOC requirements in a mod 3 pool that we use for the smallest files.
Tommy Tsui wrote:
Thanks all, that's why I am asking because there are no detail guidance on
VTOC/VTOCIX size. It always depend on ...your files no...but how can I
control the user file allocation...
On 11/22/06, Bruce Black <[EMAIL PROTECTED]> wrote:
>
> In our installation, we use VTOC(0,1,2024) INDEX(135,0,405) for mod27
This is not a big issue unless you need the space, but your choice of
405 tracks for the VTOCIX is grossly oversized. In the ICKDSF manual
there are formulas for calculating the VTOCIX size but they are awkward
since they depend on knowing the pattern of dataset names on the
volume. Similarly named datasets take up much less room in the VTOCIX
than disparate names. But anyways there is a table of the max size of
the VTOC and VTOCIX by volume size. For 3390-27 it shows 9828 for the
VTOC and 491 for the VTOCIX. These are both based on an entire volume
of single track datasets with names that don't compress well in the
VTOCIX.
Since your VTOC is only 2024 tracks, the VTOCIX only needs to be about
5% of that, or 100 tracks, and could probably be less. But there is no
penalty for making the VTOCIX too large, since it only uses the tracks
it needs to store the data it has. Extra tracks are unused.
It is the opposite in the VTOC, where every track is formatted with
Format 0 (unused) DSCBs, and many software products need to read the
entire VTOC. So oversizing the VTOC can impact the performance of
products like backup software, reporting tools, etc.
We just dealt with a customer who made every -27 the max size VTOC of
9828 tracks, but they had a good case. They found that they had many
very small datasets on these volumes and they had a number of "VTOC
full" conditions, so they just made all the VTOCs the max size.
We get calls from customers rather frequently asking this exact same
question. I think it is time for IBM to provide some more detailed
guidance on VTOC and VTOCIX sizes.
--
Bruce Black
Senior Software Developer
Innovation Data Processing
...
--
Joel C. Ewing, Fort Smith, AR [EMAIL PROTECTED]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
