number of faults in *BASE pool

[Grzegorz Goryszewski]

Hello.

I'm having QPFRADJ set to 0 , because automatic tunning had no succeed in
our production enviroment. Therefore we set
pools sizes by hand.
Here goes my *BASE pool statistics from DSPSYSSTS

Sys Pool Reserved Max ----DB----- --Non-DB--- Act- Wait-
Act-
Pool Size M Size M Act Fault Pages Fault Pages Wait Inel
Inel
2 1804,43 0,42 290 28,5 217,1 8,8 24,1 13959 0,0
0,0

Here goes *INTER pool
Sys Pool Reserved Max ----DB----- --Non-DB--- Act- Wait-
Act-
Pool Size M Size M Act Fault Pages Fault Pages Wait Inel
Inel
4 896,00 0,00 250 12,6 49,0 23,2 59,0 662,5
0,0 0,0

I know the page faults and active-wait states are way to much. My question
is how to determine amount of main memory needed
to achieve better (recommended) statistics?
Second one, how to determine optimal activity level of *BASE pool i.e. can
I substract interactive job number from all jobs showed by WRKACTJOB ?

How many faults is recommended in *BASE pool for optimal performance ?

Regards.

Grzegorz Goryszewski

[John Clark]

Hi there -

Your local IBM Business Partner should be able to help you with this.
Essentially, you can capture performance data then use BEST1 to do
modeling to figure out what the recommended upgrade path is.

I like to keep page faults in single digits (generally speaking).

Let me know if you need help locating an IBM BP. Thx, JC

"Grzegorz Goryszewski" <j2...@interia.pl> wrote in message news:<atnpde$i2g$1...@news.tpi.pl>...

[Reeve Waldock]

Tough one this with no easy answer (unless, of course, you have a glut of
poorly utilised main storage). So much depends on your total main storage,
processor size, number of pools, users, active jobs, disks etc. and then
getting the mix right for each workload. Not to mention how AS/400s /
iSeries behave at various versions against various hardware. IBM publish
guidelines for this sort of thing and I suggest you start by reading up on
work/performance management. Perhaps you could enlist the help from IBM to
point you in the right direction?

"Grzegorz Goryszewski" <j2...@interia.pl> wrote in message
news:atnpde$i2g$1...@news.tpi.pl...

[Grzegorz Goryszewski]

I don't need comprehensive answer. Just tell me if You know, what are
recommended parameters for Act-Wait and
pages fault to pages (for DB and NON-DB) , 1 ? 10 ? 100 ? , jusr
aproximatelly :). Regards.

Grzegorz Goryszewski

[Hank Knowles]

Grzegorz:

As400 tuning is really a balancing act... You have provided little
meaningful info in this post.

I can see you are running * BASE this is restrictive, and very poor
for trying to manully manage.

If you are running *base then you need to let automatic tuning do it's
magic.

What version of OS are you running, how are the disks configured, what
apps are running, IE heavy database apps on raid will tend to push
faults up....

In the old days the automatic tuning did little other than throw main
storage at the interactive pool until the rest of the system was
starved and usually choked....

I have not been on the 400 for some time but the principals still
remain the same... If you do not have the tuning or systems
experience on the AS/400 you could end up doing many IPL's getting the
system out of lock-ups.

Read the tuning manuals, or get a consultant to help you out at least
the first time through.

I applaud your attempt at fixing the issue however it maybe outside of
your individual skill set.... The 400 is tempermental and confusing,
add a poorly written app, bad communications, slow printers, over
utilized disk, etc...

I hope this may help and I am bashing you in any way shape or form.

AS/400 tuning is an art and a learned skill...

HTH

Regards,

Hank Knowles

"Reeve Waldock" <reeve....@ntlworlnospamd.com> wrote in message news:<pA7M9.6852$Om2.1...@newsfep2-win.server.ntli.net>...

[Steve Landess]

Grzegorz:

I would strongly recommend that you read (cover-to-cover) the iSeries Work
Management Guide:

See:
http://publib.boulder.ibm.com/iseries/v5r1/ic2924/info/rzahx/rzahxperfbooks.
htm

I have probably read this book cover-to-cover at least twice. Probably time
for me to read it again, too...

I'll be happy to help you if you post some additional information, like:

1) A copy of the WRKSHRPOOL display

2) A copy of the WRKSYSSTS (DSPSYSSTS) display...but please reset statistics
(F10) and let the statistics collect for approximately 5 minutes during a
busy period before capturing the WRKSYSSTS display. This time, show the
information for _all_ memory pools.

3) A copy of the WRKACTJOB display - but please reset statistics (F10) and
let the statistics collect for approximately 5 minutes during a busy period
before capturing the WRKACTJOB display.

4) Stay on the WRKACTJOB display. Place the cursor over the CPU % column
and press F16. This will resequence the active jobs into descending order
by CPU %. Capture this display.

5) Are you using QCTL or QBASE as the controlling subsystem?

6) Describe in detail the following characteristics of your subsystems,
including the controlling subsystem:
- name of subsystem
- maximum # jobs allowed in sbs
- memory pool definitions,
- job queue entries
- # jobs allowed to run simultaneously from each jobq
- routing entries
- classes used by routing entries
- Descriptions of each class (*CLS) that is used:
Class name
Run priority
Time Slice
Eligible for purge

7) Output from the DSPDSKSTS command (again, refresh(F10), then let
statistics accumulate for 5 minutes during a busy period)

8) How many interactive users do you have during peak periods?

and this is just the start for determining how to properly set up and tune
your system. However, with this information I can give you some beginning
values to use for pool size & activity levels.

Steve Landess
Austin, Texas
(512) 423-0935

"Grzegorz Goryszewski" <j2...@interia.pl> wrote in message
news:atnpde$i2g$1...@news.tpi.pl...

[Grzegorz Goryszewski]

Użytkownik "Hank Knowles" <hank.k...@wachovia.com> napisał w wiadomości
news:542556a.02121...@posting.google.com...

> Grzegorz:
>
> As400 tuning is really a balancing act... You have provided little
> meaningful info in this post.

It's because I've thought there is some tips and well know values
for activity levels and pools sizes of course dependent of numer of jobs,
maybe some mathematical formulas or so.

> I can see you are running * BASE this is restrictive, and very poor
> for trying to manully manage.

I've to manually manage pools because response time
for interactive sessions is very important, hence we set manually
size for *INTERACT pool and turned off auto tuning, because
there were problems with response time when auto tuning was turned on.

> If you are running *base then you need to let automatic tuning do it's
> magic.

I'm aware of that fact but unfortunatelly I can't do that.

> What version of OS are you running, how are the disks configured, what
> apps are running, IE heavy database apps on raid will tend to push
> faults up....

I'm on V5R1 with 1100 CPW and 240CPW for Interactive workload,
I've got 22 disk with 4 RAID5 (4 + 4 + 4 + 10) but the disks are not
the problem. We are running heavy database queries via ODBC and
ERP application.

> Read the tuning manuals, or get a consultant to help you out at least
> the first time through.

I've been told about Work Management , maybe You know something else ?

I know that tunning is kind of black magic, all I need is some tips and
highlights :). Best regards.

Grzegorz Goryszewski

[Will Burton]

I think I'm missing QSPL and QBATCH. There not showing in WRKSYSSTS, how do
I start these? Could it be that I'm still in S36 mode? I seem to remember
that there's a command to run but can remember what it was?

"Grzegorz Goryszewski" <j2...@interia.pl> wrote in message
news:atnpde$i2g$1...@news.tpi.pl...

[QMAGIC]

What do you have as your Controlling subsystem?
Check sys value QCTLSBSD.

I would think you would want QCTL, if your running QBASE then read
Work Management book to see if you want to change it to QCTL.

If your running QCTL then check your start up program and sub system
descriptions.

Dan