New ECperf Results: Oracle Achieves Best Price/Performance

Discussions

News: New ECperf Results: Oracle Achieves Best Price/Performance

  1. Oracle has tied BEA with the best Price/Performance figure of $7/BBops. They achieved a Performance figure of 19945.00 BBops/min@Std. The results were achieved running Oracle9iAS (v9.0.2.1.0) on an HP server (DL360G2), using Oracle9i (v9.2.0.1.0) DBMS, also running on an HP machine (ML570R).

    Check out Oracle's ECperf Results

    Threaded Messages (28)

  2. I assume that the test used Sun's JVM for Linux on x86.

    I wonder if Oracle will submit an ECPErf benchmark for their app server running on the Oracle JVM

    (Does Oracle still provide a JVM?)



  3. We did use the Sun JVM. The Oracle JVM is embedded in the Oracle database and is suitable for Java applications that run inside the database. Oracle9i Application Server does not use Oracle JVM in the middle tier. Oracle9iAS Containers for J2EE run on the standard Java Development Kit (JDK) virtual machine.
    Vineet Buch
    Oracle
  4. Veneet,
    Are you same person who passed 12th standard from Campion School Bhopal
  5. doesn't look like it ... he works for Oracle! :-)
  6. I wonder if Oracle will submit an ECPErf benchmark for

    > their app server running on the Oracle JVM
    > (Does Oracle still provide a JVM?)

    Well, it seems to me that EJB's should be running on the database server, and that Oracle had the perfect architecture going forward for doing just that (with the embedded Oracle JVM). However, as you can read below, they have abandoned the idea of J2EE/Corba support in the database. Businesswise, it maybe makes sense (now you have to buy an application server too), but I don't know if they really gave their own technology a fighting chance.

    From http://otn.oracle.com/tech/java/java_db/content.html

    J2EE Desupport
    Because Oracle9iAS Release 1.0.2.2 introduces a new, lighter weight, easier to use, faster, and certified J2EE container (OC4J), Oracle will no longer support the Java 2 Enterprise Edition (J2EE) and CORBA stacks in the database, starting with Oracle9i Database Release 2. However, Oracle will continue to enhance the database embedded JVM (Oracle JVM), providing Java 2 Standard Edition (J2SE) features, Java Stored Procedures, JDBC, and SQLJ in the database. ( see feature overview for more details)
  7. It's good to see Oracle weight in with their numbers.
    It is interesting to see HP being used all over too.
    Why doesn't someone use MySQL to get their performance up? ;)
  8. I wish they would simply run a test of JMS with the
    current 9iAS 9.0.2.0.0.
    We've been forced to use another JMS because Oracle's doesn't work,
    and they haven't been able to fix it yet.
    Unfortunately ECperf doesn't include JMS so these results don't give me
    any confidence it will be fixed in 9.0.2.1.0
  9. Who cares? I mean, really, is anyone looking at these results when making IT decisions? Doubtful.
  10. but Pramati still rules, right chu..mu..
  11. Who cares? I mean, really, is anyone looking at these

    > results when making IT decisions? Doubtful.

    Dear "asdf sdf",

    I believe that the ECperf benchmark is a major step forward in the appserver comparison. It is evident that you do not evaluate an appserver solely on his public ECperf numbers. These neutral numbers however do make an end at the useless wars based on vendor's own benchmarks. Although a public debate on best practices for appserver comparison may be very useful, I'm afraid this would be quite off topic here.

    But who cares about the ECperf results? I do.

    Regards,
    Pieter Van Gorp.
  12. I wish they would simply run a test of JMS with the

    current 9iAS 9.0.2.0.0.
    SPECjAppServer2002 is announced. You can ask Shanti dot Subramanyam at Sun dot COM for more about that benchmark.

    Kind regards,
    Pieter Van Gorp.
  13. Scott,
    I would be very interested in knowing what doesn't work in the Oracle JMS implementation.

    Thanks
    Andy Tael
    Oracle
  14. Andy: the problems that we've had have been reported to
    Oracle along with fairly simple code samples showing each problem.
    We originally encountered the problems, and reported them, with the version before 9.0.2.0.0,
    when we tried to port our application from WebLogic. We
    ended up using OpenJMS.
    When 9.0.2.0.0 came out, we spent a bunch more time trying to get Oracle JMS to work,
    again with no success.
    Here's a brief description of some of our problems:
    - JMS class cast exception when message selector is used in queue receiver
    - Asynchronous messaging seems to work with topics but not with queues
    - JMS messages have the same message id after they are sent
    - Persistent messages are not persisted across application server sessions
     
    If you'd like me to also send you the details, let me know.

    scott wood
    Cambian
  15. Scott:

    Agree with you ... Oracle can publish all the ECperf
    numbers they want, but at the end of the day, if it
    doesn't run my application I can't be too impressed
    with their "beebops" or "dollar per beebops".
  16. Am I correct that this is the first 100% linux benchmark? Both the app servers and the DB server ran RedHat advanced server.

    -Pete
  17. We did use the Sun JVM. The Oracle JVM is embedded in the Oracle database and is suitable for Java applications that run inside the database. Oracle9i Application Server does not use Oracle JVM in the middle tier. Oracle9iAS Containers for J2EE run on the standard Java Development Kit (JDK) virtual machine.
    Vineet Buch
    Oracle
  18. isn't http://java.sun.com/j2ee/compatibility.html a must for ECperf participation ?
  19. The compatibility with J2EE 1.2 is a must. Current version of ECPerf being about J2EE 1.2, there's no problem.

                    Yann
  20. To run ECPerf 1.0, an application server needs to be J2EE 1.2 certified (EJB 1.1 only for instance). Oracle 9 iAS is J2EE 1.2 certified.

    To run ECPerf 1.1, an application server will need to be J2EE 1.3 certified. Oracle 9 iAS is not J2EE 1.3 certified.

    I think...

    Yann
  21. J2EE 1.2 is the required J2EE level for ECperf 1.0/1.1

    As noted in ECpef 1.1 Section 4.2.2

    "The SUT must provide all application components with a runtime environment that meets the requirements of the Java 2 Platform, Enterprise Edition, Version 1.2 (J2EE) specification during the benchmark run."
  22. Hmmm.

    I question the relevance of this whole ecPerf thing.

    There are many other factors just as important as fli-flops/dollar (or whatever), such as deployability, accessibility of source, ease of use etc. etc.

    What's more when you consider there are quality appservers out there that are free (JBoss + Jonas) then the whole thing starts to sound a bit ridiculous.

    It seems to me that if you're one of these old-school IT departments that feel you really must pay 10s of thousands of dollars for an appserver then this is for you.

    The rest of us will successfully build projects for a fraction of the cost, quietly knowing the real score.




  23. I remember reading a thread not that long ago where everyone seemed to be indulging in bagging Oracle for making overstated performance claims in relation to OC4J. Personally I think a few of those words should be eaten and at least some recognition given where it is due.

    The proof of Oracle's claims is now there in open public forum BEST PRICE/PERFORMANCE AND BEST PERFORMANCE!

    At least ECPerf is a level playing field. Irrespective of how any other beloved vendor achieved thier ECPerf results, they are now dead meat in the performance stakes as Oracle has also almost doubled the ECPerf numbers posted by the nearest challenger. At least Oracle has had the guts to get down from the powerpoint podium and proove thier claims so the community should at least acknowledge that.

    Regardless of anything else, Oracle is doing positive things for Java and is completely committed to J2EE. As the largest pure-play enterprise software vendor in the world today, Oracle's committment to J2EE can only really foster positive outcomes for the J2EE community.

    Before anyone asks, "yes I do work for Oracle". From the inside, I had my doubts initially due to the history of J2EE in Oracle but frankly I think the product is now darn good and at least the equal of any other J2EE server in the market today.

    By the way, if anyone really believes that CMP 2.0 and EQL are going to provide the performance required to load, index and query millions of objects this decade think again. It took companies like Oracle 10+ years to accomplish the level of performance required to achieve this with a database server so unless I am missing something, I fail to see how multiple J2EE vendors can pull this gravity defying leap off in short order.

    So how serious should you be about J2EE 1.3 vs 1.2 right now? What do we, as developers and administrators really get out of it right now?

    Flame shields switched to maximum power...

  24. > a must for ECperf participation ?

    The release date of Oracle9iAS for this ECPERF result is August 10, 2002. So, whether or not it works right now doesn't matter w.r.t. the rules. But, presumably they must be J2EE compatible on August 10th.

  25. Guess not. Neither Oracle 9iAS release2 nor the preview release3 are J2EE 1.3 tested configurations ( aka certified ).

    From what I understand -- the scope involved in J2EE Compatibility Test Suite 1.3 ( CTS )is almost entirely different from that used in the ECPerf benchmark. See the following documentation off the ECPerf ( 1.1 ) spec :

    " .. The SUT must provide all application components with a runtime environment that meets the requirements of the Java 2 Platform, Enterprise Edition, Version 1.2 (J2EE) specification during the benchmark run.. "

    Looks like the J2EE CTS 1.2 certification will get you thru ECPerf 1.1 / 1.0 .. which means that all the pertinent stuff that makes things easier -- EJB 2.0 etc. are not part of the ECPerf suite . Which brings us to the important dilema ..The fastest possible container OR a container that runs stuff besides ECPerf ?
  26. ECperf is a database/jdbc/ejb benchmark ... there
    is little or nothing in ECperf that deals with
    the other J2EE components -- this makes it pretty
    much worthless for any real world scenario.

    IMO, it has now gotten to a state where vendors know
    exactly what to tune to get the most out of it (similar
    to the mind-boggling results one sees from TPC-C these
    days).

    Does anybody really care about "beebops" when it
    doesn't resemble anything close to reality?

    Just curious!
  27. Andy, do you work for BEA?

    Just curious!
  28. Nice try Benjamin ... :-)

    I work for the company soon to be known as "monday"!
  29. <snip>
    ECperf is a database/jdbc/ejb benchmark ... there
    is little or nothing in ECperf that deals with
    the other J2EE components -- this makes it pretty
    much worthless for any real world scenario.
    </snip>

    Not quite. If you look at any typical business app scenario, bulk of the time would be spent procesing the business logic. Further, it is here that is one is likely to hit concurrency and other performance releated issues. In the presentation layer, there cannot be much of a concurrency problem. Atleast not proportional to the EJB layer (in a J2EE app that has reasonably complex EJB components).

    In the future versions of Ecperf (SpecJAppServer*), Web components will be included too. Before that, the EJB2.0 beans will be incorporated. (In the meantime, there are other benchmarks that can assess webserver performance -even if not JSP/Servlets).

    Cheers,
    Ramesh
    - Pramati