>Given the fact that I have never seen quantititive benchmarks comparing the two, I suppose we will have to go on educated guesses here...
>
>My guess is that in-proc COM DLL's, either ST or MT, hosted by MTS would be more scaleable in the longrun. I would also guess that this scenario would be easier to manage.
>
Rick Strahl is the guy to talk to here, as he has tested it.

One understanding I have is that with MTS there is no way to limit the amount of threads that start... MTS will just merrily keep adding threads rather than Queue them up... however, I think COM+ was supposed to deal with that.

As far as performance, I think Rick said there are two issues...

1. Is instantiation... Ricks pool manager keeps the com object avilable at all times... where MTS shuts them down... now, granted MTS keeps some stuff cached... If you have hits that are close enough togther this is not an issue, but if they are less frequent there is a startup time with MTS.

2. Is the overhead of MTS... MTS is an Out-of-Process server that hosts in-process server, and there is an overhead you pay for with MTS... so, if you are not using the role based security of transaction features you are still paying for them.

I expect Rick will jump in on this thread and correct me if I am wrong. Most of what I have learned about COM, DCOM and HTTP has come from Ricks messages and white papers, and I am very gratfull to him.

BOb