JavaPro has written a roundup of the latest versions of various J2EE servers, including: Allaire's JRun, ATG Dynamo, BEA Weblogic, Oracle 9i, Borland AppServer, and Sybase EAServer. The article describes all the latest features of these servers as well as providing good backgrounhd information on the vendors and the market as a whole.

The article is at: http://www.devx.com/upload/free/Features/Javapro/2001/bgfall01/rg01bg/rg0113-1.asp

What do you guys think of their comments?

I wish people wouldn't write articles like this. It provides a bewildering array of acronyms, and for those companies in the market for app servers, being confused isn't going to promote J2EE. When I read phrases like "J2EE certified and J2EE 1.2 compliant" (Redundant), and "is J2EE certified and supports all the current J2EE (and related) APIs: EJB 1.1..." (When was this thing written?) I know that the authors just don't get what J2EE certification is meant to provide.
    The first thing I would want to know is for which J2EE standard is the server certified, or at least claim to meet. I work with this stuff every day and I can't keep the alphabet-soup and version numbers straight. Some one who knows little of the various technologies doesn't stand a chance. If I see Server A is certified J2EE 1.2, I know that an application written to that spec should work on that server. Without the J2EE spec to simplify the many J2EE technologies, neither vendors nor consumers will be able to take advantage of portability. Without it, you can forget about WORA.

I agree with John. And furthermore, the article does not (as far as I can tell) even mention WebSphere or why it is not included in the "roundup". Regardless of what you may think about WebSphere I don't think it is fair to omit it and then not even acknowledge you are omitting it.

What are ya blind? WebSphere is on the second page.

Besides i don't know why people write these articles. They're always behind the times. They cover WebSphere 4.0, but isn't it up to 4.5? They cover Sybase EAServer version 3.6.1, but its up to version 4.0 and J2EE 1.3...i mean its a waste of time even to read it.

Agreed. And the misinformation abounds. Once I got to the second page :-) I read this:

"IBM uses the WebSphere name as the branding for all of its Java development products, including VisualAge for Java, which is a JSP visual page designer that allows visual Java applet creation...".

That is news to me -- the part about VisualAge for Java being a "JSP visual page designer". Likely the reference is to WebSphere Studio, not VisualAge for Java which is a Java IDE, which is capable of running JSPs but has no such page designer.

My point is that the accuracy, never mind the versions specified -- yes WebSphere 4.0 has been out for quite a while -- leaves something to be desired and makes it very difficult to even contemplate comparing apples to apples and oranges to oranges. Just my opinion..

You're going too fast mate. The version 4.0 is brand new ;-).

    Yagiz Erkan - http://www.erkans.com
                  http://groups.yahoo.com/group/ServerSideJ

Whoops...you're right... I must be blind ;-) My apologies. I still agree this kind of roundup only serves to confuse though.

"Though independent benchmarks are not yet available (the ECPerf has not been finalized), it's likely that BEA's EJB technology would be the winner."

For those how have been busy doing ECPerf benchmarking with customers and internally this must come as a complete surprise. WebLogic has made some gains in the performance over the last few releases but I do not think that you can equate market leadership with superior performance.

Could someone from BEA tell me when they expect to publish their ECPerf results? Just confirm the authors comments.

William

Sorry to disappoint but BEA Weblogic will probably not lead the ECPerf numbers. Weblogic performs well, in my experience, but not "best". Particularly for single-server (non-clustered) application, there are faster and cheaper solutions for any one of the J2EE acronyms supported by Weblogic. What Weblogic has done well -- why I continue to use it and like it -- is that it brought all those acronyms together early on and successfully and shows balanced performance even as it scales to pretty amazing levels. Some of the fastest servers out there didn't scale at all when we tested their cluster features, but they may look good in ECPerf and work fine in single-server mode.

Cameron

Can you give some idea how much better WebLogic cluster server is compared to the others you tested? While I'm most familiar with WebLogic of all the app servers we never used the cluster feature due to the added cost. It's not clear when of if we need it.

Bob

Hi Bob,

> Can you give some idea how much better WebLogic cluster server is compared to the others you tested? <

I cannot give you up-to-date information. My comparison is about 12 months old, and at that point in time, our testing showed gains with Weblogic clustering even with the first additional server (depending on the app and the hardware etc.). Most servers performed slower with a second server in the cluster, even if "clustering features" (such as HttpSession objects) were NOT being used!

Weblogic responded better to server failure (although there were problems with failover in 5.1 that we reported to BEA and they resolved) and the cluster would show fairly balanced performance (little inter-server synchronization).

Basically, it isn't perfect, but it's good. The other servers, including some of my favorites, were just adding clustering features and they were terrible.

I expect that if you look today that your results with some of the other vendors (whom I do not want to badmouth) would be much better than my experience, since they have had more time to figure out clustering and scaling etc.

Once I took the time to get Weblogic working in a cluster -- and it did take quite some time -- I wasn't going to switch to something else, especially when nothing else to the best of my knowledge worked as well.

Weblogic has made major gains in ease of config etc. although still needs to do more. The JMS issue (queue not clustered) is a bit of a downer too, but almost no one encounters it, and you can work around it.

Peace,

Cameron.

Ah ... William the Lo(ud Mo)uth is back again!

Hey, how's BS looking? Oops, I meant BAS!

Regardless of how any app server performs on any benchmarks, it is always prudent to verify performance for your *own* application. I'm sure you know this ...

George,

Where does you bitterness towards BAS come from? Have you even used Borland AppServer? If so please provide me some details so as to determine the cause of your anger and verify your honesty.

A question, would you still hold BAS with such disregard if it came out with better ECPerf figures than the competition?

To reiterate my previous statement. BEA should be congratulated for the improvements they have made in releases this year but I do not think they will win the ECPerf benchmarking race and it is silly to assume that marketing leadership equates to technical superiority. I believe this was agreed in previous threads on this site. A fact of life would you agree.


Regards,

William

George -
Couldn't you make your statements without attacking people on a personal level?

There's a bigger wider world out there that doesn't revolve around BEA - go out and try some of the other app servers. They have many features that can be useful in projects of all sizes and budgets.


Cheers
Ray

This was just announced this morning.

News Release: Borland Takes Lead in Industry Standard Benchmarking of J2EEâ„¢ Application Servers

Industry Standard ECperfâ„¢ Enables Companies to Evaluate Standards Compliance and Performance of J2EE Application Servers

SCOTTS VALLEY, Calif. – October 29, 2001 – Borland Software Corporation (Nasdaq NM: BORL) today announced the immediate availability of the
Borland AppServerTM Kit for the ECperfTM 1.0 benchmark to enable companies to measure the real-world performance and scalability of the Borland
AppServer. ECperf is the official performance benchmark that was created through the Java Community ProcessSM program, the open community-based
process for developing JavaTM technology specifications, to test the scalability and performance of Java 2 Platform, Enterprise Edition (J2EETM) application
servers. Borland is the first company to provide an ECperf benchmark kit from its Web site at http://www.borland.com/appserver/ecperf. Today’s announcement
reinforces Borland’s technology leadership and adherence to industry standards that enable Borland customers to maximize the performance of their applications
while achieving a lower total cost of ownership.

“All users should demand their application server vendors provide official ECperf results prior to making a purchase,� said John Meyer, senior industry analyst,
Giga Information Group. “Giga recommends that organizations adopt the ECperf Specification/Benchmark Kit as the foundation for all of their J2EE application
server benchmarking efforts. In doing so, they will guarantee that their J2EE application server benchmarking results portray an accurate apples to apples
comparison.�

The ECperf Benchmark specification was developed in collaboration with industry leading J2EE application server vendors, including Borland. The benchmark
tests several key attributes of Enterprise JavaBeansTM (EJBTM) technology, including scalability and distributed transactions. By requiring that the source code
cannot be changed, the ECperf Benchmark specification takes advantage of the portability of Java and J2EE technologies and provides a rigorous comparison of
J2EE-based server performance. It is a powerful test benchmark of the scalability and performance of J2EE application servers.

“We’re pleased to see Borland assume a leadership role in providing customers a kit for ECperf to easily and quickly determine the performance of the Borland
AppServer,� said Rick Saleta, group marketing manager, Enterprise Java Technologies, Sun Microsystems, Inc. “We encourage other vendors to provide their
customers with a Kit for ECperf. ECperf benchmark based testing will help users make an informed purchase decision that will subsequently maximize their return
on investment and will best fit their specific enterprise needs.�

The primary goal of the ECperf Benchmark is to model the workload of a real-world system that would include manufacturing, supply chain management, and
order/inventory. The ECperf Benchmark specification was created to test the performance of business objects in the middle tier – a critical determining factor
when choosing a server for the best price performance for current and future business demands. The Borland AppServer Kit for ECperf 1.0 Benchmark is being
provided for customers to test for themselves the scalability and performance available from Borland AppServer. The Kit features a step-by-step guide for
customers to install, run, and analyze ECperf results for Borland AppServer.

“Companies who want to gain the best performance for their investment can turn to Borland for a high performance application server at the lowest total cost of
ownership,� said Alan Shoap, vice president of Enterprise Solutions at Borland. “In fact, we encourage companies to use the ECperf benchmark to test the
performance of Borland AppServer. Since the ECperf results for many highly scaled applications can vary dramatically, Borland also encourages companies to
compare benchmarks and request analytical tools from other vendors.�

A whitepaper entitled the “Eight Reasons ECperf is the Right way to Evaluate J2EE Performance� has been developed by Sun Microsystems and CustomWare.
It is another useful tool for customers to evaluate J2EE application servers using the ECperf Benchmark. The whitepaper is available for download from
http://www.customware.com.
Borland also announced today that it has joined TheServerSide.com partner program, dedicated to providing a J2EE community web site and the official
repository for ECperf results, available at https://ecperf.theserverside.com/ecperf/. Borland expects to publish results of the Borland AppServer ECperf
benchmark during the fourth quarter on both www.theserverside.com and www.borland.com.

Have you all taken wind of this thread?
https://www.theserverside.com/home/thread.jsp?thread_id=9906#32358

Does this report include jBoss?