Re: Remote outq vs. *DEVD

Energy saver is a factor. The guy who tends to these printers is
looking into that as well. Problem is that one area has four of these
printers. Really handy as nobody has to get up and walk to a printer.
It also means that they will go to energy saver mode.

The spool file contains plain text lines and is produced by RPG. The
spool file gets sent directly to the printer where the internal forms
chip does the work, after printer wakes up, of course. InfoPrint is
not used. No overlays, either. No IPDS, plain text only. The issue
of installed printer memory was brought up as well. I think they have
96M of memory. All this to print a single duplex page.

There is a very short customization object involved. It tells the
printer to go into postscript mode and run the forms program.

No forms definition or font substitution messages. Very rare form
change messages are trapped by reply list entries. Usually they only
happen when writer is started.

I do not believe that we are anywhere near to the max active of 33.
Hospital administration would be elated if there was that much
activity. At best, we >might< have ten printers getting activity at
any one time.

I can almost guarantee the latest print group PTF, along with cumes
have not been applied. I can't understand the magic power the sa has
to avoid dealing with this. I sneaked in the install of PTFs
suggested by Scott K to fix QSH error reporting. Got away with it
because no IPL was needed. During upgrade to v5r4, the only cume was
installed, years ago.

I did a traceroute from my pc and the i to one of the "slow" printers.
Pc was uninformative, reporting less than 1 ms. The i, however,
reported .760, and that remained the same even when I changed the port
to use 9100 explicitly. I had wondered if the network was an issue.
A total of two hops, whether from my pc or the i, to the printer.
Longest time was from printer itself, and that is the point 760 time.

John McKee

On Tue, Dec 27, 2011 at 11:24 PM, Jim Franz <franz400@xxxxxxxxxxxx> wrote:

John,
A laundry list of possibilities:

With several hundred lasers, & max active is 33, depending upon how often
they have spools to print, could be too low.

I would also check on the hardware side, and see if these printers going
into an
energy saver mode that needs time to wake up.

As Larry pointed out the biggest issue has usually been limited memory in
spool subsystem.

Is this a spooled file generated from rpg or similar, or is this spool sent
to a monitored
queue, where some "package" software processes it, and outputs a 2nd
spooled file that actually gets sent to the printer outq? If so, the
subsystem
that runs the package software may be memory or max active constrained.
Is InfoPrint Server involved?

Are the Lexmark Optra's with IPDS cards?  & how much memory in the printer &
ipds card?

I know you said the lasers have a form on a chip, but the spooled file may
have other resource issues:
Is the final spooled file attributes:
Printer device type ??
Device Requirements - what is flagged as "Y" ??
Front and or Back overlays?
Form Definition?
User defined objects or data?

Any font substitution messages on qsysopr?

Isn't performance hunting fun??
The redbooks IBM eServer iSeries Printing VI and VII, IBM AS/400 Printing V,
have many references to performance issues. Printing VI and VII have a lot
to do with InfoPrint client & server, but also other printing issues.
Printing V has a lot of AFP and PSF/400 issues, but also other printing
issues...

Finally - make sure the latest Print Group PTF for your OS version is
installed, and if InfoPrint Server (I could not find a list of ptf's but I'm
sure there is one...)

Jim Franz

----- Original Message -----
From: "jmmckee flinthills.com" <jmmckee@xxxxxxxxxxxxxx>
To: "Midrange Systems Technical Discussion" <midrange-l@xxxxxxxxxxxx>
Sent: Tuesday, December 27, 2011 8:37 PM
Subject: Re: Remote outq vs. *DEVD


We have several hundred laser printers attached.  I don't know the
percentage that are set us as remote output queues.  Most of the time,
very low print volume.  Does the number of output queues effect
performance even when no data is being sent to be printed?

Also, I heard a comment attributed to a third party.  The comment, as
best as I heard it and can remember, (noisy room), was that it could
take 90 seconds for a queue to get rescanned for output to print, and
performance depended on when new output arrived in the queue.  That
sounds odd, somehow.  I have no idea where the 90 seconds came from.
As to CPU load, this box has never appeared slow.  The few times I
have looked, percent CPU has never been even over 30.

Makes me wonder if spooling subsystem priority is also a factor.  But,
I am loathe to mess with settings and get the sa who resides in his
own world, "offended".  As far as I can tell, this issue should really
be his to deal with.  But, nobody wants to get involved with him.
Where can I get a job where nobody ever asks me to do anything and I
don't get fired for doing almost nothing?  Very sad situation.

John McKee

On Tue, Dec 27, 2011 at 3:45 PM, DrFranken <midrange@xxxxxxxxxxxx> wrote:

I suspect very little. If each is using host print transform they both
need to do that work and the rest is communications.

First thing I would take a look at is the size of the memory pool used
for spooling/printing. Bring up WRKSYSSTS and use F21 to set to 2 or
3(Intermediate or Advanced). Use F10 to reset statistics then have
someone crank up a print job that is tagged as 'slow'. Using F5 (Not
F10) refresh the screen and watch the Fault columns for the *SPOOL
pool. I have seen those numbers go to 4 digits then I see the spool
pool is at a measly 4 or 8M....... and that's the problem. If that's the
case the pool needs to hold a bit more memory when it's not being used
so that when work comes along it has space to work.

To fix it use WRKSHRPOOL and adjust the 'Minimum Size %' column for the
*SPOOL pool. (You'll need to press F11 to see it) The default is 1
meaning it can shrink to 1% of the total memory on the system. For a 4G
machine that's 40M but if you have a 1G machine it's only 10 which isn't
likely enough. Available memory is displayed at the top of that page.
Using that, figure out the percentage needed to set the 'floor' at about
32M to start. Then test again to see if performance improves. If not
perhaps 48M is a good next step.

Of course if the machine is pounding along at 100% CPU this might not
help as much so standard warning about YMMV!

I have seen this small change turn 2 minutes into 10 seconds for a print
job, hopefully it helps you as much!

- Larry "DrFranken" Bolhuis

--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at http://archive.midrange.com/midrange-l.