Well, if they are all near the max, half track blocking would result
in 2 blocks of 2 records.  A 32K block size would result in a block of
3 records, with space for another block of 2 records, but I'm not sure
QSAM will do that.  And 2.4GB is about 90% of a 3390-3.  An average of
7000 or less would be great with 1/2 track blocking, but even the
minimum size of 2K for a very simple record you could even just go
without blocking without losing much space to inter record gaps.

How about a pattern of some sort on the DSNs.
HLQ7.aaaaaaaa.bbbbbbbb.ccccccc.Dyymmdd.Thhmmss, etc.

On Thu, Oct 8, 2020 at 11:08 AM Michael Stein <[email protected]> wrote:
>
> On Wed, Oct 07, 2020 at 07:41:58AM -0400, Joseph Reichman wrote:
> > The average number of records in the file could be 240,000 each record
> > could be close to 10,000 bytes there are 4,644 files
>
> So each file could be up to: 10000 * 240000 -> 2.4 GB
>
> And the maximum total data is: 2.4 GB * 4644 -> 11.14 TB
>
> Reading 11 TB in 4 hours (?) is:
>
>   11e12/(4*3600.) -> 763 MB/second
>
> Most likely all the records are less than the maximum (10K bytes) size
> and there is less data that this.
>
> That's a lot of records too: 240000*4644 -> 1114 million records.
>
> How long does a billion (1000 million) QSAM GET locates take?
>
> I have no idea how long a QSAM GET takes.  But a billion of them
> add up and it's all CPU time.
>
> What's the blocksize of these datasets?  The CPU for each GET
> is just bumping a pointer and checking for the end of the current
> block.  There's more CPU involved in switching blocks.
>
> I'm still wondering:
>
>   Is this a one time job or recurring every year, month, week, day?
>
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions,
> send email to [email protected] with the message: INFO IBM-MAIN



-- 
Mike A Schwab, Springfield IL USA
Where do Forest Rangers go to get away from it all?

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

Reply via email to