BizTalk goes to sleep

[Riaan]

This is a repost from awile ago, I did not get any responses - is it
maybe because nobody knows or the issue cannot be bypassed/corrected:

It appears as if my BizTalk server goes to sleep. The statement sounds
funny, but this is my conclusion based on the following: I have a
BizTalk server with a two-way NetTCP receive port. My application has
an orchestration that subscribes to the receive port. The orchestation
then calls a NetTCP web service. The response is sent back to the
original service consumer. I use the orchestartion to do basic message
enrichment and other required tasks.
I have tuned the MaxReceiveInterval and batch size settings in the
adm_serviceclass table to optimize for low latency.
(MaxReceiveInterval = 50, BatchSize = 1).

The following behavious happens:
Application A calls the NetTCP service receive port with a request.
The request is services in 3 seconds. The same request is the called
again, right after the previous request. The request is returned in
500ms. Every subsequent request after this is also around 500ms, given
then they are executed with almost no wait time between them.

Wait for approx 30 seconds, then Application A issues another request.
The request is exactly the same as previous request, but it then takes
3-5 seconds.

My deduction is then if as certain amount of time goes be between
activity on the port (or msgbox database), some thread goes to sleep
and needs to be woken up again on the next request.

I am currently keeping the solution "awake" by having a ping request,
which is a small request being issued against the port. This is not
ideal as it is skewing my stats and generating unneccessary traffic.

Is there anyone that has come across this and knows how to fix this?
Is there some sort of a setting or something that will override this
default behaviour? And what exactly is this behaviour, why is biztalk
doing this?

Riaan Gouws

[Francois PIETTE]

Hi !

Is the computer under heavy load ? Does it have enough RAM ?
A possibility is that the code to process the request is swapped out
probably because of other activities. Use PerfMon to check memory swapping,
disk I/O and other parameters. Perfmon draw a graph so you can see the
impact when you send your request.

--
Francois Piette
http://www.overbyte.be

"Riaan" <r.gou...@gmail.com> a �crit dans le message de news:
01f5235c-dc68-409a...@s31g2000yqs.googlegroups.com...

[Riaan]

Hi

Both the BizTalk 2006 R2 server and the database server is idle. There
is no other activity on both servers. The SQL server has 8GB of ram
and the BizTalk server has 4GB of ram and both servers have ample RAM
available - no swapping should be happening.

Riaan

[Dan Rosanova]

I would use perfmon to check the throttling performance counters.
This can happen a lot and cause exactly what you're seeing.

Kind Regards,
-Dan