Uday,
Lots of things could be having an impact here. Can you give us a little
more information?

What AS/400 are you running on? What other workloads also run on that
AS/400.

Is WAS allocated its own memory pool? How big is that pool?

Are you using JDBC? If so, are you using the Toolbox driver or the
Native driver? And are you using a DataSource? And are you using
Prepared Statments?

There is WebSphere performance information here:
http://publib.boulder.ibm.com/infoce...v6r0/index.jsp


Walker.

We are using iSeries V5R3 and installed WebSphere 6.0 Express Edition
also configured datasource using JDBC Toolbox driver V8 to connect to
remote iSeries DB.

Yes. We are using Prepared Statements.

The same web application is deployed on Windows Server and it is
working very fine.

I have monitored hit ratio using WKSSYSSTS, WRKDSKSTS, WRKACTJOB and
coudn't find any CPU utilization and ASP utilization but non-db faults
is quite bit large.

We are using default memory pool for websphere server.

Can you guide me anything tuning parameters is required to set on
iSeries to improve performance? Also any parameters need to be setup on
WebSphere Configuration?

Regards
Uday

The first two things I would do are:
1) Isolate WebSphere in its own fixed-size memory pool (I don't know
what the default pool is for 6.0, but for 5.0 it runs in *BASE by
default, which is hopeless from a performance point of view). This is
crucial. The exact size of this pool will depend on your application,
but 1Gb would be a good place to start (you may need to allocate more).

2) Use the Native JDBC driver. If WebSphere is running on an AS/400 and
talking to an AS/400 database you can use the native driver (even if
WebSphere and the database are on different AS/400s). You should only
use the Toolbox driver when WebSphere is not running on an AS/400, but
talking to an AS/400 database.

There are other steps you can take, but it is difficult to offer more
advice without answers to the other questions in my first post: What
model and processor AS/400 are you running on? (e.g. Model 820,
Processor 2395) What other workloads also run on that AS/400?

Have you looked at this website?:
http://publib.boulder.ibm.com/infoce...c400/index.jsp

Walker.

iSeries Model i520 and having 2GB Memory and 1000CPW.

Actually IBM has recommended to use IBM Toolbox Driver. We have taken
PMR support from IBM during datasource configuration and we coudn't get
any connection error in log using above driver also transaction is
working fine.

WRKDSKSTS is also showing 35% disk utilization. I guess this would be
fine.

I have checked that CPU utilization is not more. But there are non-db
page faults available. We are using *BASE pool for WAS.

Pool Memory Max Act Non-db Pages
faults
*MACHINE 180 91 .1 1
*BASE 2200 200 24 124
*QINTER 550 70 3 5

Websphere process using 64% Thread and total number of thread is 180.

I tried to increase MAx Act levels and also *BASE Memory pool but still
page faults is not decreasing.

Can you please guide me about iseries performance issues.

I tried to monitor using wrksyssts, wrkactjob, wrkdsksts but coudn't
improve the peroformance.

I guess it running very slow due to using maximum thread on iseries.
Can you please guide me about threading configuration on iseries? Where
we can set the maximum/minimum threads?

Thanks for considering my request.
Uday



walker.l2 wrote:
> The first two things I would do are:
> 1) Isolate WebSphere in its own fixed-size memory pool (I don't know
> what the default pool is for 6.0, but for 5.0 it runs in *BASE by
> default, which is hopeless from a performance point of view). This is
> crucial. The exact size of this pool will depend on your application,
> but 1Gb would be a good place to start (you may need to allocate more).
>
> 2) Use the Native JDBC driver. If WebSphere is running on an AS/400 and
> talking to an AS/400 database you can use the native driver (even if
> WebSphere and the database are on different AS/400s). You should only
> use the Toolbox driver when WebSphere is not running on an AS/400, but
> talking to an AS/400 database.
>
> There are other steps you can take, but it is difficult to offer more
> advice without answers to the other questions in my first post: What
> model and processor AS/400 are you running on? (e.g. Model 820,
> Processor 2395) What other workloads also run on that AS/400?
>
> Have you looked at this website?:
> http://publib.boulder.ibm.com/infoce...c400/index.jsp
>
> Walker.

Hi

iSeries Model : i520
We have created 2 LPAR on above i520 model.

1st LPAR - HTTP and WebSphere Express 6.0 on OS/400 installed.
300 CPW, 1.5 GB RAM

2nd LPAR - DB2/400 on OS/400 installed.
700 CPW, 3.5 GB RAM

We have deployed web applications in WebSphere on iSeries.
The Web Site is running just below acceptable speed now after I did
below mentioned Performance Tuning.

I have done tuning on AS/400 with respective to the subsystems and as
well for the WAS Specific.

1. I have already done some Performance tuning on iSeries using
following parameters

Increased the *BASE Memory pool (Default memory pool size was 1850 and
now I have set it to 2700)

Increased *Base Pool Thread Activity level (Default was 140 and now I
have set it to 200)

When I raised to above 200 the website is slow.

I have SET *NOMAX for QMAXACTLVL (Maximum activity level of systems)

Set Java Running Thread Priority (Default was 20 and now I have set it
to 65)

Please guide me any other settings is required to improve Performance?


2. Tuning Parameters set for a WebSphere and also for the Java
Environment

JAVA Virtual Machine Heap Size - Initial Heap Size 256 MB and Maximum
*NO Max (0)
Servlet Engine thread pool size - 70
Data Source Connection Pool - Min 40 and Max 65 Connections
Web Container Thread Settings - Min 50 and Max 60 Connections

By setting the above parameters the performance has improved to some
extent.
I would like you to review this information and please guide me about
how to use Tivoli Performance Viewer to monitor thread/connections/JVM
and any other parameters require during concurrent users accessing web
site.

Hope above information will help you to guide me on right track.

Thank you.
uday





uskulkarni wrote:
> iSeries Model i520 and having 2GB Memory and 1000CPW.
>
> Actually IBM has recommended to use IBM Toolbox Driver. We have taken
> PMR support from IBM during datasource configuration and we coudn't get
> any connection error in log using above driver also transaction is
> working fine.
>
> WRKDSKSTS is also showing 35% disk utilization. I guess this would be
> fine.
>
> I have checked that CPU utilization is not more. But there are non-db
> page faults available. We are using *BASE pool for WAS.
>
> Pool Memory Max Act Non-db Pages
> faults
> *MACHINE 180 91 .1 1
> *BASE 2200 200 24 124
> *QINTER 550 70 3 5
>
> Websphere process using 64% Thread and total number of thread is 180.
>
> I tried to increase MAx Act levels and also *BASE Memory pool but still
> page faults is not decreasing.
>
> Can you please guide me about iseries performance issues.
>
> I tried to monitor using wrksyssts, wrkactjob, wrkdsksts but coudn't
> improve the peroformance.
>
> I guess it running very slow due to using maximum thread on iseries.
> Can you please guide me about threading configuration on iseries? Where
> we can set the maximum/minimum threads?
>
> Thanks for considering my request.
> Uday
>
>
>
> walker.l2 wrote:
> > The first two things I would do are:
> > 1) Isolate WebSphere in its own fixed-size memory pool (I don't know
> > what the default pool is for 6.0, but for 5.0 it runs in *BASE by
> > default, which is hopeless from a performance point of view). This is
> > crucial. The exact size of this pool will depend on your application,
> > but 1Gb would be a good place to start (you may need to allocate more).
> >
> > 2) Use the Native JDBC driver. If WebSphere is running on an AS/400 and
> > talking to an AS/400 database you can use the native driver (even if
> > WebSphere and the database are on different AS/400s). You should only
> > use the Toolbox driver when WebSphere is not running on an AS/400, but
> > talking to an AS/400 database.
> >
> > There are other steps you can take, but it is difficult to offer more
> > advice without answers to the other questions in my first post: What
> > model and processor AS/400 are you running on? (e.g. Model 820,
> > Processor 2395) What other workloads also run on that AS/400?
> >
> > Have you looked at this website?:
> > http://publib.boulder.ibm.com/infoce...c400/index.jsp
> >
> > Walker.

Uday,
Another website you might want to take a look at it is
http://publib.boulder.ibm.com/infoce...ex.jsp?lang=en
specifically you might be interested in the performance section.

Now to come to your specific points:

Your disk activity seems to be at a reasonable level. But I wonder how
much % of your disk space is still free on the AS400?

I must admit I'm very surprised IBM have advised you to use the Toolbox
driver, I've always found the Native driver to have much better
performance, but maybe there is some issue with WAS Express 6.0 under
v5r3 that means it can't be used.

I'm puzzled by the memory figures you quote: first you said that you
had 2G, but the WRKSYSSTS figures show nearer 3G; then in your
partitioned setup you have 5G allocated, but you say you have increased
*BASE from 1.8G to 2.7G when you have only allocated 1.5G to that
partition. How mcuh physical RAM do you have? And are you allowing the
memory to be automatically moved between partitions?

You really shouldn't run WAS in *BASE, but in its own memory pool
instead. By partitioning the 400, you appear to have achieved the same
effect (isolating WAS memory use from any other memory use).

Page faulting occurs when data has to be fetched from disk rather than
RAM, therefore you will always see some when your application starts
(because obviously the AS400 hasn't had a chance to load it into memory
yet). The Max Act level only has an indirect impact on page faulting.
The amount of memory allocated can have an impact - if there is not
enough RAM for your application the disk will be used instead (not good
for WAS performance).

In your LPAR setup, I would add at least another 100CPW to the WAS
partition, 300 CPW is a little on the low side; or are you allowing the
CPWs to be moved between partitions automatically?.

Changing the Java thread priority from 20 to 65 has made those jobs a
lower priority. Is this what you wanted to do? (It might be, if your
performance problems were due to slow HTTP running rather than slow WAS
running.)

As far as the WAS parameters go, I really can't say whether they are
the best for you because I don't know your application. However, they
are not unreasonable settings, so probably aren't causing you any
problems. What is your PreparedStatement cache size though?

I haven't used the Tivoli stuff much, and we run WAS 5.0 Base, so I
suspect what I have done will not really be relevant to your setup.
Have you got the Runtime Perfomance Advisor switched on? It should give
guidance in the System.out file if things are not set up well.

What level of logging do you have set for the IBM Service Logs? I think
by default it is set to Log All, if you change this to Log Warnings &
Errors, or to Log Errors, you should see a performance improvement.

How long do you leave the WAS servers running on the AS400 before
stopping and restarting them? In WAS 5.0 by default the JIT compiler
won't kick in until a method has been called 2000 times, up until that
point they are interpreted (this threshold is adjusted by a JVM
property ). So if you are cycling your WAS servers every day, you might
find that you are always running interpreted code; but if you leave
your WAS servers up for weeks on end, you should find that the
performance gradually improves as more and more methods get compiled.

Walker.

Hi

I have got the exact details about iSeries configuration.

iSeries Model : i520. We have created 2 LPAR on above i520 model.

1st LPAR - HTTP and WebSphere Express 6.0 on OS/400 installed. 500
CPW, 3 GB RAM
2nd LPAR - DB2/400 on OS/400 installed. 500 CPW, 3 GB RAM

I have revert back to Java Thread Priority from 65 to 20.

#### As far as the WAS parameters go, I really can't say whether they
are the best for you because I don't know your application. However,
they are not unreasonable settings, so probably aren't causing you any
problems.

----- Our application is mainly using for online billing (ebpp). User
can view online bill (pdf) and pay the bill online. database write/read
activity is randomly. Can you please give me your inputs on WAS
parameters, atleast I can experiment and let u know best feedback. I
hope it will benefit for our application also.

#### What is your PreparedStatement cache size though?
- PreparedStatement cache size tis 10.

I have enabled Runtime Perfomance Advisor using websphere admin
console. I will look into systemout.log during peak hours.

We have stoping and restarting WAS Server at many times during testing
the application and Since last 3 weeks appserver is not stop.

Thanks

uday

Uday,
Your PreparedStatement cache size seems small. If you have more than 10
SQL statements in your application, you should increase the cache size.
IBM recommend that you don't make the cache bigger than 1,000 though.

Walker.