Keep forgetting that changed.

So to amend what I said, the best minimum in my example would be BUFNO=8
(2 128k Large Blocks or 1 256k large block)

Darren

-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On
Behalf Of Ed Gould
Sent: Wednesday, February 13, 2008 3:44 PM
To: [email protected]
Subject: Re: 3590 stacking performance via JCL?

On Feb 13, 2008, at 3:44 PM, GAVIN Darren * OPS EAS wrote:

> Most of the Performance on tape drives can be accomplished by matching
> the BUFNO= parameter to the Devices I/O Data Transfer Size. For your
> model that's probably (most Tapes and VTS are) a 128kb size, so you
> would want enough buffering to hold at least 1 128k Large Block.
>
> So for a BLKSIZE=32768 file you would want a BUFNO=4 at minimum, but
> would be better off with BUFNO=16 or BUFNO=32 (4 or 8 Large Blocks) as
> Some VTS Units and the 256 track 3590E have a 256kb Transfer Size.
>
> However DFSORT ignores any buffering (It has its own I/O context) on
> both DASD and TAPE so no need to use it when doing sorts.
>
> The other thing you can do is to add a step at start of job that will
> Pre-Mount the tapes and then Retain them for any following steps,  
> using
> IEFBR14 as a stub program, and the VOL=(,RETAIN) option mentioned
> earlier.
>
> There is also the /*setup card that can be used on some JES systems.
>
> Darren
>
>

Darren,

I am fairly sure that the default for QSAM  is 5 buffers if you  
decrease it to 4 you will have a negative impact on performance.

Ed

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

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

Reply via email to