Send async message between bpm processes takes long time

[Former Member]

Hi all,

We use XI 3.0 SP16.

We have 2 BPM Processes.

One of these Processes sends an asynchronous message, which is directed

to the other (by the receiver determination defined in the directory).

In SXMB_MONI, we see that the "End Time" of the asynchronous send step

is about 3 seconds after the "Start Time".

Furthermore, when the first process sends this message to several other

processes, this time is longer, and gets up to 10 seconds for the

last "send" step in SXMB_MONI.

What can be the cause for this?

How can we minimize this time?

Thanks ahead for any help,

Yoav.

[STALANKI · Accepted solution]

can u please tell me what is the need for a BPM for ur scenario?Can u give the scenario and end system landscapes?

You can try performance tuning guide from servce market place.!

[Former Member · Accepted solution]

Hi Sravya,

Thanks for your answer.

The main process is similar to the Multicast pattern.

The other process(es) is an Async/Sync Bridge.

We use A/S Bridge for every receiver so that we can first send all requests asynchronously,

and only then handle responses as they arrive.

(Otherwise, a request is sent only after the former response arrives).

However, paradoxically, using A/S bridge makes the whole process take longer (for few receivers) due to the problem described.

Yoav.

[andreas_stolz]

Hello Yoav,

keep in mind that you see the ccBPM processing time as well. ccBPM is based on workflow, the message need to be copied into container variables. So there is some work to do.

The entry queue into one ccBPM process is EOIO. If you send mutliple messages into one process, they're queued and wait until the predecessors are finished. That might explain why the processing time goes up to 10s.

What about the general load on your system?

Rgds.,

Andreas

[Former Member]

Hi Andreas,

I hava one "Container Operation" step after the "Receive" step.

In that operation I add a message to a multiline message of the same type.

What is the "normal" time a step like this should take,

for a message of size 100kb?

Regarding the load on system:

This is a development system so load is very low.

However, we did not perform XI fine-tuning yet.

Yoav.

Message was edited by: Yoav Gur

[Former Member]

Hi

One of the problem for slower performance of BPM may be lots of workflows there in cache.Goto swwl TCode and delete all the workflows and try to execute it again.

Also goto sxmb_adm and try to activate the queue.

Hope this works out.

Regards

Victoria

[Former Member]

Hi victoria,

Clearing swwl and activating the queue, helped a bit.

But still, the sending of the message takes few seconds.

In addition, I guess it is not feasible to clear the swwl on regular basis in a production system.

Anymore thoughts?

Yoav.