Hi all,
this may be weird question but I want to know how many maximum adapters or different insatnces of adapter can run peacefully on single IS box.
Is there any therotical way to decide this or it has to be thorugh trial and error?

Thanks
Neal

Neal,

Great question. I think the answer will depend on two primary factors:

1. The configuration of the server in terms of processor(s), memory, access to fast disk systems and operating system [*]The actual processing load of the integration services using the resource adapters.

Simply hosting the adapters doesn’t consume much in the way of resources. Processing load is created when you execute Flow (or java) services that invoke adapter services for those hosted adapters.

You could have 100 adapter instances with no problems if you never used them. You could also run into performance problems with a single adapter if you had a underpowered server, incredibly high transaction volumes or a (very) poorly architected solution.

Reminds me of an old commercial jingle “So, how many licks does it take to get to the chocolate center of a Tootsie Roll Tootsie Pop?” Your question has the same answer. It depends.

HTH,

Mark

Thanks for responding Mark. Ya I know it is million dollar question
But is there any way if I can come with some data that will say I have x number of adapter instances and y number of services will run simulteniously using them with avrage document size of z. I nedd so many IS instances. Or it has to happen through trial and error only?

thanks
Neal

There are calculations in some of the performance tuning documents that help you determine the amount of memory (heap size) and the size of the IS thread pool based on document size, documents interarrival times, number of triggers, etc.

However, those calculations will not tell you how many adapter instances you can fit in a theoretical integration server.

I’ve always had better luck using the available sizing calculations as a rough starting point for a server configuration and then verifying those assumptions through actual performance testing and tuning than with relying on some spreadsheet developed by someone you don’t know for a purpose that might not match your own situation very well.

Mark

Mark,
thanks for your inputs.

Neal