Objects by Design Forums
Show 15 posts from this thread on one page

Objects by Design Forums (http://forums.objectsbydesign.com/index.php)
- Object-Oriented Software Environments (http://forums.objectsbydesign.com/forumdisplay.php?forumid=10)
-- J2EE vs .NET controversy (http://forums.objectsbydesign.com/showthread.php?threadid=545)

Posted by SZ on 11-04-2002 02:41 AM:

J2EE vs .NET controversy

It looks like The Middleware Company has embroiled themselves in a bit of a controversy by publishing benchmarks comparing J2EE to .NET that show considerable performance advantages for .NET over J2EE.

However, it turns out that the open forums of their sister company, The Server Side, have exposed a huge amount of skepticism about the way the Java Pet Store application was implemented in J2EE as the basis for comparison with an equivalent .NET implementation. To read the details of this brewing controversy, see the links at the top of the page link given above, especially the FAQ and the messages on the Server Side forum thread for this benchmark.

To summarize, it appears that:

1) It was too expensive to rewrite the Java Pet Store application from scratch to optimize it for J2EE and therefore, EJB may have been used in a sub-optimal way for an all-out speed contest.

2) New standards such as CMP in EJB 2.0 were not available in the application servers tested, although it may be a better model to use for this benchmark.

3) The JVM's for Linux on an Intel 32-bit architecture don't necessarily reflect what can be done on 64-bit Solaris today.

4) The bottom line is that Microsoft is a single vendor solution so it can be tuned all the way through. However, the benchmark reveals the fragmentation in the Java market due to multiple not-always-compatible vendors (read the report for details).

5) A rematch is brewing. See the links at the top of the page given above. The rematch will involve the participation of the Java expert community on The Server Side to a much greater extent.

From our standpoint, we would be happy to see the experts who work with the open-source application server JBoss 3.0, which fully supports EJB 2.0, do an optimized Java Pet Store implementation that would be fully open-source. If these folks, who know the J2EE platform inside out can't produce a good benchmark for comparison with .NET then we are afraid the Java community will have a lot to justify in the near future when competing head-to-head with MS .NET in the marketplace for enterprise-scale application server middleware.

Posted by SZ on 11-20-2002 02:51 AM:

It turns out that there is more to the story than meets the eye.

Rickard Öberg has written a first-class critique of the The Petstore Revisited benchmark at the following page: http://dreambean.com/petstore.html .

This is one of the better analyses that we have seen of a benchmark in a while. In his analysis, Rickard tears apart many of the claims made by The Middleware Company (TMC) with regard to "optimizing" the original Java Petstore code.

Moreover, this critique raises some serious credibility questions about TMC's motivations. In particular, it appears that TMC had funding and cooperation from Microsoft to do this benchmark. See the critique for full details. This is amazing. Very soon on the heals of the Microsoft settlement with the Justice Department...

As far as web services goes (a part of TMC's benchmark), a friend reported that he had very good results with Apache Axis 1.0, a Java open-source implementation of the W3C SOAP standard. He reported that he was able to connect to Axis from Visual Studio .NET with full SOAP, WSDL and UDDI functionality.

So you see, the race has just begun...

Posted by SZ on 11-22-2002 02:42 AM:

To understand EJB 2.0 and CMP - container-managed persistence - read the excellent article by Richard Monson-Haefel at IBM's developerworks: Read all about EJB 2.0.

Richard Monson-Haefel is a co-founder of the OpenEJB open-source project and the author of Enterprise Java Beans, the popular O'Reilly book.

The new EJB persistence model is a lot cleaner than the 1.1 persistence model. Also, the persistence management is now delegated to a persistence manager, which may be from a different vendor than that of the EJB container. Why is this important? Because now the database specialists out there can provide highly optimized persistence solutions.

The TMC Java Petstore benchmark will look like a joke once somebody rewrites it using the best that the latest version of J2EE has to offer.

Posted by sduskis on 11-26-2002 02:45 PM:

OBD forum callenge

The TMC Java Petstore benchmark will look like a joke once somebody rewrites it using the best that the latest version of J2EE has to offer.

Is this a challenge to the OBD community? This sounds like a good opportunity .

Posted by SZ on 11-26-2002 11:59 PM:

A suggestion was made on the TSS forums to host the rewrite of the Java Petstore benchmark on sourceforge.net. Since they host JBoss and OpenEJB, it is probably the best place for this to happen, and we would love to see those two open-source EJB containers used in the benchmark.

Imagine the publicity for open-source software!

Posted by SZ on 12-11-2002 03:08 AM:

See Sun's website for a new best-practices example of how to use J2EE called Adventure Builder 1.0.

Perhaps Sun tired of the PetStore example and went back to the drawing board. Benchmarks anybody?

Posted by sduskis on 12-11-2002 02:16 PM:

SZ, it looks like Adventure Builder will be the blueprint for J2EE web services. This may be a better starting point for performance comparisons between J2EE and .Net.

For now this means that we'll have a tutorial for creating J2EE web services. That's a nice move on Sun's part.

Posted by felchlin on 01-11-2003 05:48 AM:

Lightbulb P.S. It's not only about performance

Another aspect of the Petstore controversy is that the criticisms of J2EE focus on performance. An equally relevant goal of J2EE (and Java) is the vendor-independence and open, cross-platform compatability aspect. Any J2EE server and database may be used and interchanged. Microsoft will never compete on that level.

Full disclosure: I am an avid Java programmer, and I despise Microsoft and what it stands for.


Posted by sduskis on 01-13-2003 05:16 PM:

Rickard Öberg, who wrote the Review of "PetStore revisited", has some very interesting things to say on his blog (web log) about the latest article from Javaworld that discusses the controversy.

The basics of his blog:

  1. "If they had asked J2EE vendors for participation they would have vetoed the whole thing."
  2. There was no conclusion about performance from the test, even on the windows platform.
  3. .Net encourages a "style of writing code [that] does *not* scale to large projects"

Rickard is one of the key players in this controversy. I find it worthwhile keeping an eye on his work on this controversy and his other endevours.

Posted by prash03in on 02-04-2004 10:52 AM:

Complete solution

I have worked for a lot of clients now and have seen that all the big names in finance/supply-chain and consumer goods industry in US and Europe are now going for a vendor specific supply of the entire environment rather than just a server or a database.

What I mean to say is, if this is the case, don't you guys think that MS has an edge here? MS has a great tuning of the entire environment, the big clients now, as they are going for vendor specific complete environment, what will the J2EE community have to say.

According to me, the most important feature of Java was platform independence and also, I have seen people pointing out that the J2EE will work with equal ease if we change the server or the backend.

My question, why do we need to change the environment? The J2EE and java code might work on all the servers and backends, but is it as fast as MS .NET even on a single platform???


Posted by SZ on 02-05-2004 08:08 PM:

Take a look at the Linux 2.6 benchmarks on 64 bit processors in the latest issue of Infoworld. Also, the new multithread support (NPTL) in the 2.6 kernel now provides the Java JVM with incredible scalability, which is needed for J2EE application servers.

So you see, the competition has barely begun.

All times are GMT. The time now is 08:02 AM.
Show 15 posts from this thread on one page

Powered by: vBulletin Version 2.3.5
Copyright © Jelsoft Enterprises Limited 2000 - 2018.
Copyright ©1999-2005, Objects by Design, Inc.