V12.2 throughput overhead

[Former Member]

Hi,

We have a 12.2 system running well except for a presumed throughput limit. An underlying query, depending upon filter criteria, may contain many thousands of records. We have seen, and can replicate, that for smaller record sets, everything within MII works fine, and we can call the record set via a standard SQL query and display it in a standard iGrid on a standard IRPT page, but when the record set gets to a certain size, it (the web page) simply goes in to la-la land.

System is MII V12.2.2 on NW CE 7.1 EHP1 SPS 5 with AIX DB2

Some quantification;

1. If I run a large query in SQL directly (stored proc which is accessing a separate database), it takes approx 1 sec to return the data set completely. But when I run the SAME stored proc via fixed query in the MII workbench it takes around 30 secs to return data.

2. Each record by my calculation is around 158 bytes, and we can replicate la-la land at around 8,000 records. As I mentioned, smaller record sets work fine.

Obviously when I run a large query from an IRPT, which is attempting to populate an iGrid, the web page does not respond, and the "Updated" event is hence never triggered.

Any ideas?

I'm quite intrigued as to why in the the workbench we are seeing so much overhead. There are no errors, just a LOOOOOOONG wait for the data to be fed through to MII.

Note that I can, from a separate SQL server, query the source database and retrieve 100,000 records in < 2 secs, so the pipe to the database seems fine.

Thanks

Kevin.

[jcgood25]

Hi Kevin - some troubleshooting ideas:

1. Check the NW logs to make sure you aren't getting timeout errors or anything that would point to root cause for http threads acting up.

2. Add the <PARAM NAME="Trace" VALUE="true"> to the iGrid applet and watch the java console to see if anything relevant shows up (check the java console to start with in case you are getting an out of memory error on the client side - good thing to not set the RowCount excessively high either).

[Former Member]

Hi Jeremy,

I've limited the row count to something that appears to be manageable. For larger record sets, a web page error usually appears as : "Target applet or JVM process exited abruptly".

Can't report anything else specific at this stage, except that my main question still remains, which is that even without an applet or web page, just in the workbench, a query that takes 4 seconds to execute in the database directly takes 120 seconds to execute in the workbench.

Regards

Edited by: Kevin Hibbins on Mar 5, 2012 4:01 AM

Edited by: Kevin Hibbins on Mar 5, 2012 4:02 AM