Adavance Adapter Engine Performance Tuning

Question

Hi all,
We are using PI-7.11 .  We are working on an asynchronus scenario that reads JMS queue and  calls  server proxy at the receiver. The interface is configured  to bypass the ABAP stack(Integration server) by  using advance adapter configuration. The PI-7.11 supports Proxy through a soap apter at the receiver. All this works fine and problem we are facing is the message processing time. Currently we are getting  a throughput of 10 messages per second and each message of size 2k.  If we combine 100 such messages into a single message then we get a throughput of 50  single messages per second. To keep up the incoming messages on the sender queue, we need to process at least 30 messages per sec. Instead of packaging(combing) the messages , we  created a couple of communication channels to read the same JMS queue and simulated the multi-threading on reading the sender queue. The reason we are doing this because we do not  want write an intermediate process to package the messages from the source system.  The bottle neck we are seeing in  the RWB(Run time work bench) is the 'DispatchDisp' queue  in  the adapter engine. Based on the RWB display, this queue has single thread  and it can not keep up  with the multiple senders .  Anyone knows how this queue being used in the Advance adapter engine and how can we change this to multiple threads.  Is it used before calling the PI mapping  or after calling the mapping ? We are able to increase the number of consumer threads of other queues related to the JMS  queue in the adapter engine.
Thanks in advance.
Meghanth

Answer 1

Hi Meghanth,
From what i hear it sounds like you can push more into your adapter engine with the JMS sender then your SOAP receiver can process in the same time, right?
The DispatchDisp Thread (1) is usually not the bottle neck, as far as i know the Dispatcher is just dedicated to dispatch messages to the receiver threads.
If you see an accumulation there (i guess in Message Display Tool->Engine Status under Queues) its a typical sign, that your receiver Threads are all busy and cannot accept more at that time.
How many Threads did you assign to your SOAP Receiver?
Usually the possible count for throughput with the available Threads is easy to calculate that way:
E.g. transfer to your ECC takes 200 ms (check your audit log steps) -> you can process about 5 / second / Thread ..with e.g. 5 Threads maximum 25/sec).
You can try analyze this with the "old" MessagingSystem Monitoring, there are timestamps in milliseconds available. (http://<host>:<port>/MessagingSystem).
According to that i d try to improve the count of your possible receiver threats..(take care that there are enought threads overall on your machine and the receiver ECC can take this).
With best regards
                  Sebastian