Our objects have transactional settings which means that they need
synchronisation on. However, if we could remove the transactional settings
and turn synchronisation off, would this mean that we have to perform our
own thread safety mechamisms? We do not have mulithreaded clients and we
don't share object instances between clients/other objects, so would this be
necessary? Will COM+ internally share object instances in some way if
synchronisation is off, so we are forced to do our own thread safety?
The threading model of our objects is "Both".

Charles.

-----Original Message-----
From: arun [mailto:akumarOMNESYSINDIA.COM]
Sent: 06 October 2001 06:55
To: DCOMDISCUSS.MICROSOFT.COM
Subject: Re: FW: COM+ and Free Threading

hi Charles

I have the same task coming up regarding porting of NT4.0/MTS2.0 setup to
2000/COM+ and I have been following your discussions closely albeit i really
am not
very much up about how you measured the performance of your calls. How are
you doing this ? Using Mike Nelsons performance sample (dt.zip) or COMPERF
that comes with MSDN or are you using some external tools?.
And how have you been measuring the time spent in the database with respect
to your calls.
Has anyone of you used reationals performance suite to do this work before,
and know
if it is suited to this task?

arun

----- Original Message -----
From: "Charles Gamble" <Charles.GambleSINGULARITY.CO.UK>
To: <DCOMDISCUSS.MICROSOFT.COM>
Sent: Thursday, October 04, 2001 9:56 PM
Subject: Re: FW: COM+ and Free Threading

> Hi Steve,
>
> The reason we started running this test was because we had noticed a huge
> performance drop for our product when it was used on Windows2000/COM+.
> We found that without changing the architecture of our system, there was a
> 2.5 times drop in performance between MTS and COM+.
> The test mimics our architecture except without the DB work, etc. which
the
> methods calls perform. Note, our system in written in C++ and does
database
> access using ODBC.
> SQL Server is the DBMS. Object1 is a controller DLL, Object2 is used to
> create transactions where necessary, Objects 3 and 4 are business objects.
>
> The fact that you may need to setup Object3 and Object4 into an
> un-configured library application to get closer to the same performance as
> NT seems to suggest that you cannot use the same architecture in COM+ as
MTS
> and expect the same performance. In fact, most suggestions that I have
heard
> so far seem to suggest removing the number of configured components in
your
> app. According to Tim Ewalds book, Transactional COM+, he states that
using
> library apps will get improvements in performance but this is basically
> setting them up as raw COM objects albeit they can exist in the callers
> context. Surely, this is the whole point of MTS/COM+, in that, you can
> configure all your components and let everything be taken care of
> (synchronisation, transactions, etc.) for you.
>
> We ran that test using your script on the library package and I got a 200
> millisecond speed increase. The COMTIInstrinsics and IISInsrincsics
settings
> seem to bring the performance gain. So now its 1360 milliseconds in COM+
> compared to 631 milliseconds on NT.
>
> Charles.
>
>

----------------------------------------------------------------
Users Guide http://discuss.microsoft.com/archives/mailfaq.asp
contains important info. Save time, search the archives at
http://discuss.microsoft.com/archives/index.html .
To unsubscribe, mailto:DCOM-signoff-requestDISCUSS.MICROSOFT.COM

----------------------------------------------------------------
Users Guide http://discuss.microsoft.com/archives/mailfaq.asp
contains important info. Save time, search the archives at
http://discuss.microsoft.com/archives/index.html .
To unsubscribe, mailto:DCOM-signoff-requestDISCUSS.MICROSOFT.COM

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]