Re: wrksyssts non DB-fault paging

["Steve Landess" <sjl_123@xxxxxxxxxxx>]

> "Steve Landess" wrote:
> subject: Re: wrksyssts non DB-fault paging
> ...
> 1) Why do you have over 45,000 jobs in the system?
> You're probably keeping too many spooled files around...
> Remember, the system has to keep up with these jobs even after they have
> ended.

>> Then Marty Urbanek wrote:
>> Are you serious, Steve? "Keep up with" in the sense of *actively* keep up
>> with, i.e. keep checking on, i.e. using CPU cycles?
>>
>> This could be a problem for me, too. We have tons of spool files sitting
>> around, mainly because it is hard to automate cleanup of our individual
>> users' stuff because they have legitimate reasons to keep some things for
>> long periods of time. Certainly though, the vast majority of it is
clutter,
>> because of not enforcing an automated cleanup.
>>
>> I just thoght it was wasting some disk space. If the system is actually
>> doing work to maintain this clutter, I need to invest a little more time
>> into a cleanup strategy.
>>
>> What kind of system resources are consumed by these old jobs' spool files
>> laying around?


Marty,
It has been a while since I did a lot of performance tuning, but the answer
is an unqualified *yes*.  However, I can't tell you exactly how much
CPU time is spent keeping up with these old jobs.

Bottom Line:
You should work to eliminate unnecessary spooled files from the system
(such as old job logs) so that these jobs will disappear from the system and
thus the WRKSYSTS screen.

You should also probably IPL the system at least once a month
to clean up the job tables.

See the following quote from the Work Management Guide
(SC41-5306-03), page 128:

<Snip>

Job Tables
The operating system uses internal job tables to track all jobs on the
system. Each
entry in the job table contains information about one job.

Job Table Entries
When a new job enters the system, an entry must be available for the new
job. If
there are no entries available in the table for the new job, the table is
extended.

Your system will experience a performance degradation when job tables are
extended. Too many available entries will lower your system performance
during the
IPL steps that process the table and during runtime functions that work with
jobs.

If the number of available entries is large, you can use the Change IPL
Attributes
(CHGIPLA) command to change the option to compress the jobs tables. For more
information about the CHGIPLA command, see the CL Reference (Abridged)

</Snip>

As an aside, you can ameliorate the slowdowns that occur during job table
extension
by changing some of the system values related to jobs:

QACTJOB      *ALC      Initial number of active jobs
QADLACTJ     *ALC      Additional number of active jobs
QADLTOTJ     *ALC      Additional number of total jobs
QMAXJOB      *ALC      Maximum number of jobs
QTOTJOB      *ALC      Initial total number of jobs


You should print out the Work Management Guide and read it from cover to
cover.
I find this book to be very helpful in understanding the nuts and bolts of
performance
management:
http://publib.boulder.ibm.com/iseries/v5r1/ic2924/books/c4153063.pdf


I would also suggest that you take a look at the Performance Management
Redbook:
http://publib-b.boulder.ibm.com/Redbooks.nsf/RedbookAbstracts/gg243723.html?Open

>From Chapter 9 (System Performance Tuning Tips) of the Performance
Management
Redbook, under the topic "Operating Environment Tips":

(the "guidelines" are found in the Work Management Guide)

<Snip>
Interactive  AS/400  Tuning  Roadmap

1.  Enter  command  WRKSYSSTS

2.  Wait  2-3  minutes  and  press  PF5  to  refresh

3.  Does  *MACHINE  NDB  faults  meet  the  guidelines?
    a.   Yes  ...  Press  PF10  and  go  to  step  4.
    b.   No  ....  Adjust  QMCHPOOL
        1)  -50K  if  fault  rate  =  0
        2)  +50K  if  fault  rate  >  3.0
        3)  Press  PF10  to  reset  and  go  to  step  2

4.  Is  the  DB  fault  +NDB  fault  >  20  in  any  pool?
    a.   Yes  ...  Increase  pool  size  by  50KB,  press  PF10  and  repeat
step  4  (repeat  until  all  pools  are  less  than  20)
    b.   No  ....  Go  to  step  5

5.  Wait  2-5  minutes,  press  PF5.  Is  the  Wait  to  Ineligible  state
=  0  ?
    a.   Yes  ...  Reduce  Activity  level  by  2,  press  PF10  to  reset
and repeat  step  5
    b.   No  ....  Go  to  step  6

6.  Is  the  Active  to  Wait  state  10x  the  activity  level  ?
    a.   No  ....System  not  heavily  used  or  complex  application  mix
go  to  step  4
    b.   Yes  ...  Go  to  step  7

7.  Is  the  sum  of  all  fault  rates  for  all  pools  within
guidelines?
    a.   No  ....  Go  to  step  4
    b.   Yes  ...  Go  to  step  8

8.  Activity  levels  and  pool  sizes  probably  OK.    Continue
monitoring WRKSYSSTS  display  regularly.

Figure   31.    AS/400  Tuning  Roadmap


9.12   Operating  Environment  Tips

IPL your System: Occasional  IPLs  are  no  longer  required  on  the
AS/400  for optimum  operation.  However,  it  is  at  IPL  time  that
permanent  and  temporary addresses  are  regenerated,  permanent  and
temporary  job  structures  are  built and  so  on.   Refer  to  AS/400
Operator's  Guide  for  more  information  but  try  to  IPL at  least  once
a  month.

Note  that  temporary  job  structures  are  built  during  run  time  as
needed, according  to  system  values  QADLACTJ  and  QADLTOTJ  described
earlier  in  9.4, "Review  System  Values"  on  page   143.

</Snip>


Regards,
Steve Landess
Austin, Texas
(512) 423-0935