A recent Oracle-run benchmark pitted the three application servers against each other on Sun Microsystem's Java Pet Store Demo to test the average response time of each server with 600 users. The result was Oracle 9iAS running twice as fast as Weblogic or Websphere.

read more:
http://www.oracle.com/features/9i/index.html?9iasfast.html

Yada, yada, yada...
Where are Oracle's ECperf results???

I agree! I've been involved in doing performance tests on many of the hyped appservers, of which all boast on performance etc, etc, etc compared to other vendors appservers... and boy have I been lied to!

Thank God for ECPerf! At last something that the vendors can actually PROVE how great they are.

ORACLE has an ORACLE benchmark that shows it has a faster app server than IBM or BEA?? Can you say "sales pitch"? I thought you could!

:p

Well done Orion!!!... oops sorry I meant Oracle ;)

The same debate keeps on repeating..
If ORACLE really wants to show their product ask them to come with their ECPerf results. I guess it is not worth considering any performance metrics with Java Pet Store !!!

I would have to agree. I have used
WebSphere and Oracle / Orion ( both distro's ).
I like them both but show me the ECPerf results.

Ok, maybe it's not the write place to say this...
I REALLY would like to see a real comparison between an aplication running .net and an aplication running j2ee, in any app server. Not the bullshit MS has released as an article where she implemented the pet store, but a real fair comparison, done by a non-biased third party. Will this ever be possible?

J2EE & .Net. Ok that is totally laughable.

How can u expect two different technologies to release a benchmark on which they can agrees upon, when J2EE vendors running a standard technology cannot agree on the same.

ECPerf has been here for long. but who cares, i mean the vendors. beleive me after one app server releases comparison of ecperf results, the other vendor will come up with something to throw it right back at the other vendor.

do u think, they will take ecperf as the last benchmark performance test on this earth, if they r defeated. hey its a mutli-million dollar industry. everyone wants a share.

I´ll put it simple.
I have a real world application, say a pet store. I want to see how it scale when I use j2ee, and also when I use .net. Both technologies allow the implementation of such app, so it´s simple.

Of course, when I say "scale", I mean raw performance only, i.e., how it performs as long as the number of clients increase.

I´m also not considering manutenability, oop compatibility, nothing. I want to see a raw throughput of each platform.
I don´t thing it´s laughable... The corporate managers, as a rule of thumb, want to see THIS first, and then the other stuff(manutenability, standards compatibility, etc.).

Andre Augusto

"I have a real world application, say a pet store. I want to see how it scale when I use j2ee, and also when I use .net. Both technologies allow the implementation of such app, so it´s simple."

No it isn't. For example, the .NET application that Microsoft wrote as a Pet Store benchmark dropped some key maintainability features. It doesn't use MVC and has no separation between the application logic and the display code. That is not a reasonable approach for a real e-commerce site that has to worry about internationalization, co-branding, separation of tasks (Java developers vs. HTML coders), etc. I would never allow a design like that in a system I was responsible for.

"Of course, when I say "scale", I mean raw performance only, i.e., how it performs as long as the number of clients increase."

In other words, you mean nothing very useful. If all you care about is raw performance, why don't you code it in assembly language? And write your own OS? You'll definitely beat the pants off of .NET and J2EE. For that matter, you'd probably beat both of them with a tightly coded app in C or even PHP. Choosing things like J2EE has more to do with maintenance and interop with other systems than it does with simple performance.

"I´m also not considering manutenability, oop compatibility, nothing. I want to see a raw throughput of each platform. I don´t thing it´s laughable..."

And that's where you're wrong. It is laughable to worry about raw performance (which can be improved by purchasing additional hardware) with no consideration for maintainability, compatibility, or overall cost.

Hello Perrin,

I´d like to thank you for your answer. I think you really got the point. I know everithing you put, and I´m a long time advocate of all of this. All I wanted was to satisfy a curiosity, but, as you show, it´s out of logic. Ok.
(Sorry for my english - manutenability... I´m still learning... :) )

Andre

What's funny about this stuff, is how seriously such big guys like Oracle and MS take Pet Store. The main purpose of this app is to show to j2ee newbies as much j2ee guidelines as possible. It's buggy and unstable. A half of a year ago we reported to Sun two bugs in one single servlet, that prevented Pet Store from being accessed by more than one client.

This is not a real world application. And it's funny just to think about somebody using Pet Store as a benchmark. I think Oracle is missing something. Decision making has been shifted from developers to managers of IT departments. They don't make decisons based on performance of "Pet" app.

Again you may laugh but I'd like a proper comparision between this app on J2ee and the .Net which is supposedly 28 times faster.

While 'it's not a proper comparison' most end web users don't care about what's under the scenes (and quite frankly neither does business if they can give users what they want)and if some 'perfect' OO derived proper J2EE 'reference' implementation is beaten hands down by a MS product, I'm going to find it hard to persuade the COO why we're not using .NET! And what sort of a 'reference' implementation is this anyway, if it gets beaten by several other companies who write it in a different way (???), Oracle and MS included? Come on Sun - give us something written well!

Keeping your head in the sand about this doesn't help people like me who work in mixed environments. I know this was done with C# but who cares if it's not done in a 'proper oo' language (e.g. VB) or however it's done - if you can use it to beat J2ee Performance hands down in less lines of code (supposedly). Does someone want to answer me that seriously rather than just pretend MS doesn't exist?

PetStore is designed for concept proof. It uses all J2EE technologies for available of J2EE today.

By the way, Oracle uses its own way to implement J2EE. For EJB, Oracle uses ORMI instead of RMI/IIOP. Although it is a thin and faster, I don't think it will be adopted by most of companies because it is not the standard. The others modification include use stored procedure to implement most of business logic and then lost the extensibility of J2EE. Absolutely it will be faster.

It is great for Oracle do a lot for Java. It happens same as Oracle's database version 6. We'd better wait for its next version to proof its story to be true.

Haven't we seen this discussion / theme before ??

Oracle 9i and JDeveloper ROCKS!

I kind of notice that Oracle app server is Orion app server. Is this true? What is the relationship between these two company? Does anyone know?

Thanks

Oracle licensed the souce code from Ironflare. Their current versions are pretty much the same, but if Oracle is going to add stuff to the code base, or fork off, remains to be seen.
Last I heard was that they are doing some minor work together, like bug-squashing, but thats the only relationship between the companies to my knowledge.

We've been through this before ... from my personal experience, this isn't anywhere close to the truth. Oracle was the slowest when I tried the very same benchmark!

Been there, done that! Let's move on ...

"Oracle claims it takes only 30 minutes to demonstrate that Oracle9i Application Server runs 2-4 times faster than BEA WebLogic Server. BEA accepted the challenge and found that, contrary to Oracle’s claims, BEA WebLogic Server is at least 54% faster than Oracle 9iAS."

http://www.bea.com/press/releases/2001/0709_benchmark_bea-ibm-oracle.shtml

I think we've seen this release before since Oracle doesn't say when their "recent" becnhmark was carried out.

Regards