Parallel execution of business process steps

[jpenninkhof]

In PI 7.1 we have noticed that all steps in a business process are executed by BPE in a serial way. I.e. there won't be more than one concurrent thread used for the execution of a business process.

In our business process we query many systems (approx. 5~10) and aggregate the data in an answer to a consumer. Because these systems all take their own sweet time to respond, you can imagine that the entire process takes quite a while the be executed. In our scenario, a user is waiting for the response and is amused with the time he needs to wait.

We would like to see the business process to be split into several concurrent threads, so that the various systems are queried in parallel rather than one after each other, which would hopefully give us a significant performance boost.

I was told that PI 7.1 doesn't have the capability of splitting a business process into several concurrent threads, but I'm afraid a "cannot" is not an acceptable answer to our users. So I would like to know if anyone has insight in:

1. Secret possibilities that may be less obvious, but still present in PI 7.1 that would allow a business process to be split up in several threads
2. Whether the new PI/PO 7.3 has a capability to split a business process into several concurrent threads, or if this feature is on the roadmap
3. If there might be add-ons that can be installed on PI/PO that would allow the break-up of a process into several concurrent threads
4. If none of the above works, what your opinion is on how this should be solved. Should we e.g. re-build the process in Java-code, where we would have more control over the creation of threads? Perhaps use another broker specifically for this purpose? Any idea would be welcome!

Thanks in advance, experts!

[Shabarish_Nair]

have you tried the multiple queue option in the BPM?

Do read section 4.2.3.3 Queue Assignment of this document http://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/e0715b1d-68a6-2b10-9f8f-86c5b059c...

[jpenninkhof]

Thanks for your answer, but this is related to the inbound queues of PI. Our challenge is that we have one message going into the BP from where 4 systems are called using a fork.

The fork seems to be serial instead of parallel, so we need some kind of mechanism that really calls the 4 systems in parallel and then gathers the answers into one response.

At the moment each system takes about 3-4s to respond, meaning that our total response time is already about 15-16s and we still need to add more systems.

Any help would be much appreciated.