Originally posted by: SystemAdmin
We use one agent on a standalone server. That agent manages all of our HTTP traffic used in DataStage TX. Initially we wanted to install it as part of our existing web cluster just in case one box failed another would pick it up. We finally had a Mercator consultant come in to try configuring it within the cluster. Spent a week, couldn't get it to work with 6.5.2. We may have had some factors that complicated things though. Installed it on a stand alone no problems. (If clustering is crittical you might want to look at RTI. I don't know anything about it but I'd think it would be a required feature.)
At this time the agent is running on a 500 Mhz desktop with 250MB ram and a small HDD. We aren't using the box for anything else. DataStage TX is installed on it's own, separate, server.
Once you start testing there are some settings in the mercator.ini, DataStage TX, file that may help performance. Also, within each msd I create two instances of the watch.
jvanboga
PS: "...but your experience seems to be positif. "
{depends on who you talk to. Our web developers think they can write a faster app. We load tested internally with a Grinder script. Our tests hit the agent with anywhere from 5 - 100 simutanious queries recycling the hits every 3 seconds for 25 - 75 cycles. They thought that performace was poor. We had a failure rate of well below 1%.
Our national association tested 1 - 3 simultanious hits every 3 - 5 seconds for a specified number of cycles. The only failures were anticipated.
Currently we average only 300 hits per day (300 queries + 300 responses) and haven't had any problems with the Agent. Tansactions are usually under 1kb each way. Most activity occurs in a 10 hour window. So far, its been more stable than the DataStage TX server which handles all our EDI activity as well. blah, blah, BLAH.....}
#IBMSterlingTransformationExtender#DataExchange#IBM-Websphere-Transformation-Extender