Originally posted by: SystemAdmin

We've successfully migrated numerous maps in our Sun Solaris env to Resource Registry. However, we've run into a performance issue in production with a recent auditing map deploy. This auditing map runs for each map run, so has heavy use. The audit map with 2 db connections hard coded can process 60-100K logs/hr. With RR, we get a max of 5-30K logs/hr, although we also see apparent competition for rr access, as audit bogs down when other maps are accessing rr.

We've had the best success with dtx.ini MaxThreads at 20. We have the audit map on a separate launcher.

Suggestions welcome.

Develop in wtx 8.1.

Thx.
J
#IBM-Websphere-Transformation-Extender
#DataExchange
#IBMSterlingTransformationExtender

Originally posted by: repanzer

The resource registry allows use to resource varibles under launcher in the event an resource for a map is different under deployment server then is it for another.
Such as, on one box, the input is;
c:\Input\*.txt

and the same map when deployed to another box, should have an input of;
c:\InputData\*.txt
This way, you can update the map and deploy to both servers at any time, and you don't have to switch;

c:\Input\*.txt

to

c:\InputData\*.txt

when you deploy.

That being said, I have no idea what you mean by;
"We've successfully migrated numerous maps in our Sun Solaris env to Resource Registry
#IBM-Websphere-Transformation-Extender
#DataExchange
#IBMSterlingTransformationExtender

Originally posted by: SystemAdmin

Our primary use of resource registry variables is for database connections, passwords, and queue names. This allows the same code to be portable across our different regions. We have recently changed (migrated) existing maps to remove mdq references and queue names and replace them with resource registry variables. We have been happy with the performance of those maps in production.

We have been unpleasantly surprised by the drop in prod performance with resource registry settings compared to mdq connection in a heavily used auditing map. There are only two locations in the map that have changed from mdq to RR, but the performance drop is significant. We like the code portability, but we need to minimize the performance hit.

Any suggestions on ini settings or theories on what could be causing the drop are appreciated. All maps use the same mrn file. Is there a limit on the mrn?

Thanks.
#IBMSterlingTransformationExtender
#IBM-Websphere-Transformation-Extender
#DataExchange

Originally posted by: repanzer

I got ya.
I had never used the RR before my current shop, and my current shop abused the heck out of using the RR. Even if an input path was the same string for every deploy server, my team mates used the RR for it, in which there is no technical reason to build and use the file if the value is the same.

Because the RR was used from the get go, I wouldn't have noticed an increase or decrease in processing time.

As far as I know, there is no limit to how many mrn's are used in the mrc.

The only thing I can think of, and this is only if there are a lot of entries in the mrn, is to make sure the ones that are used the often, maybe the DB info, should be the first entry in the mrn file, and that MRN file be the first file listed in the mrc.

Any time map sees a resource registry entry, it goes to the mrc, determines which server the map is running on, then goes through each mrn file, and each entry in those files, looking for a match. If you have 10 mrns listed in the mrc, and the mrns's have lots of entries, and if the DB connection info is the last entry in the last file, then making it the first entry in the first file MAY increase the speed. Or maybe putting the DB connections in their own MRN file and have that mrn listed first in the mrc.

and make sure to change all the mrc settings; global, command server, and launcher, are pointing to the same mrn's.
#IBMSterlingTransformationExtender
#DataExchange
#IBM-Websphere-Transformation-Extender