Welcome to MSDN Blogs Sign in | Join | Help

Welcome to The Metaverse

Navigating the service-oriented, identity aware metaverse

News

  • Disclaimer:
    The content of this blog are my own personal opinions and do not necessarily represent Microsoft's position, commitments or strategy. In addition, my thoughts and opinions often change, and as a weblog is intended to provide a semi-permanent point in time snapshot you should not consider out of date posts to reflect my current thoughts and opinions.




    Add to Technorati Favorites
Enterprise Services performance paper published!

Many people are concerned about the performance of their Enterprise Services components compared to their old COM+ implementations. In order to discuss this issue, I (along with a couple of colleagues) wrote a paper published here on MSDN.

The paper essentially highlights the most optimal way to create Enterprise Services components and that if our guidelines are followed then your Enterprise Services components will, by and large, match the performance of COM+ components!

Comments welcome ;)

Posted: Monday, March 29, 2004 4:59 PM by RichTurner666

Comments

Robert Hurlbut's .Net Blog said:

# March 29, 2004 8:11 PM

Timothy said:

"This clearly illustrates that the impact of COM+ transactions is negligible in these tests."

I always believed that using the DTC results in a certain amount of overhead compared to ADO.NET/SQL transactions, let alone no transactions.

Or am I missing something?
# March 29, 2004 11:52 PM

Rich Turner said:

There is a certain amount of overhead involved in executing a distributed transaction, but in our tests, the transaction overhead was negligible compared to the amount of work being performed.

Also note that the tests were executed on servers with fast hard drives, multiple processors and lots of RAM - typical of a modern server environment. If the tests are run on less capable machinery then the results will be different and the cost of the transaction will increase as a percentage of the overall test. This is particularly the case for machines with slower HDD subsystems because they will be working harder or waiting longer for slower hard drives.

HTH.
# March 30, 2004 9:41 AM

John St. Clair said:

Rich,

Thanks for the nice article. I'd like to see a few additional topics addressed however:

1. I'm not sure that the JITA perf test is correct. Sure, it works if you want to compare the cost of a single method call, but for those cases in which multiple method calls are made on an object, I think the results are likely to show that turning JITA/deactivating after each call would actually hurt perf.

2. Services without components: although this is .NET 1.1/Windows 2003-only feature, this would be interesting to me. I've started looking into this, and think it would be an interesting comparison to add to the mix. I have no idea on performance, but assume that in many cases it would be slower; I'd love to know, however, that I'm wrong!

Thanks again,
John
# March 30, 2004 11:30 PM

Rich Turner said:

Remember that performance and scalability are orthogonal. Yes, in certain configurations, calling a JITA component multiple times may be slower than calling an object to whom you hold a reference. But this perf' hit is likely to be (and should be) insignificant compared to the cost of doing work. However, systems using JITA components are likely to be FAR more scalable than non-JITA components.
# April 5, 2004 5:10 PM
Anonymous comments are disabled
Page view tracker