Sun submits SPECjAppServer2004 score on free stack

Home

News: Sun submits SPECjAppServer2004 score on free stack

  1. Sun recently submitted a SPECjAppServer2004 submission based on a software stack that's entirely no-cost: Solaris 10, MySQL, and their platform edition app server.

    The SPECjAppServer2004 application is an industry benchmark for testing Java Enterprise Edition applications. It's designed around an application that has features in five domains. From the design document:
    The five SPECjAppServer2004 domains modeled are : Dealer, Manufacturing, Supplier, Customer, and Corporate. The Dealer domain encompasses the user interface components of a Web-based application used to access the services provided by the other domains (i.e., Manufacturing, Supplier, Customer and Corporate), which can be considered "business domains." In a company of the size being modeled, it is assumed that each business domain, has separate databases and applications. Most likely, they are implemented on separate computing hardware also. There are producer-consumer relationships between domains in the company and to outside suppliers and customers as well. For historical reasons, each business domain may have distinct entity IDs (i.e., the customer ID used in the customer domain may be different from the ID used for that same customer in the Supplier domain). It is for this reason that global customer or supplier databases are likely to exist in the Corporate domain, with accompanying applications.
    With three server nodes and one database instance, it scored 266 JOPS (jAppServer Operations Per Second), calculated as Dealer Transactions/sec + Workorders/sec. (For reference, the SPECjAppServer2004 FAQ mentions the application being run on 1.6GHz laptop with a 30GB HD, with 512MB RAM, and while the DB isn't specified, it attained a score of 2.) While this is the lowest score reported, comparing the configuration to the other submitted scores and factoring in the costs of software licensing, this score is quite good.

    Note, also, that the nature of the application tested is "large," and not the most popular application of Java, Enterprise Edition; most applications developed tend to be much lighter and therefore might generate far different transaction patterns and stress points. Another factor is that the SPECjAppServer2004 application is designed to work with J2EE 1.3, not J2EE 1.4 or JEE 5, so some of the solutions might be considered "archaic" by the current specifications.

    It might be interesting to see if there is potential in creating a standard "typical" JEE application, to generate transaction comparisons for non-Fortune 500 applications.

    Threaded Messages (21)

  2. I'm good to see Sun doing this.

    Am I the only one dying to know what the scores would be if the identical test was repeated, but Sun swapped out the Sun app server for Jboss, Geronimo and Caucho Resin? (Other open source servers.)

    Peace,

    Cameron Purdy
    Tangosol Coherence: Clustered Shared Memory for Java
  3. I'm good to see Sun doing this.Am I the only one dying to know what the scores would be if the identical test was repeated, but Sun swapped out the Sun app server for Jboss, Geronimo and Caucho Resin? (Other open source servers.)

    No, you're not.
  4. I'm good to see Sun doing this.Am I the only one dying to know what the scores would be if the identical test was repeated, but Sun swapped out the Sun app server for Jboss, Geronimo and Caucho Resin? (Other open source servers.)
    No, you're not.

    Me Three.

    I have been using the SunApp server with mixed results. In general I really like it, but on the other hand it sometimes craps out, and starts throwing stack traces that don't seem related to my code.

    I'm also happy about the good integration with Netbeans. I think it will end up benefiting the community to have a well integrated IDE/App server for those who want it.
  5. Me Three.I have been using the SunApp server with mixed results. In general I really like it, but on the other hand it sometimes craps out, and starts throwing stack traces that don't seem related to my code.

    Matt, if you didn't have any luck via the forums :

    http://forum.java.sun.com/forum.jspa?forumID=136

    shoot me an email with some details (sharps at sun dot com) and I'll see if I can help.

    Rich Sharples
    Sun Microsystems
    http://blogs.sun.com/roller/page/sharps
  6. Me Three.I have been using the SunApp server with mixed results. In general I really like it, but on the other hand it sometimes craps out, and starts throwing stack traces that don't seem related to my code.
    Matt, if you didn't have any luck via the forums :http://forum.java.sun.com/forum.jspa?forumID=136shoot me an email with some details (sharps at sun dot com) and I'll see if I can help.Rich SharplesSun Microsystemshttp://blogs.sun.com/roller/page/sharps

    Thanks Rich,

    I took a break from working with it last month, but I plan to pick it back up. Thanks for the link to the forum, for some reason I missed it last time I went looking for help.

    Any word on when you guys will have a EJB3 beta version of platform edition out for us to try?
  7. Thanks for the link to the forum, for some reason I missed it last time I went looking for help.Any word on when you guys will have a EJB3 beta version of platform edition out for us to try?

    It's taking shape - specifically some bits of the EJB 3.0 container are already in GlassFish and more is being added over the next couple of months. Some details on Milestone 2 here :

    http://glassfish.dev.java.net/servlets/NewsItemView?newsItemID=2399
     
    You'll probably hear about it first on the GlassFish project page - there's a feed here :

    http://glassfish.dev.java.net/servlets/ProjectNewsRSS

    or you can subscribe to the "announce" mail alias here :

    http://glassfish.dev.java.net/servlets/ProjectMailingListList



    Rich Sharples
    Sun Microsystems
    http://blogs.sun.com/roller/page/sharps


    Rich Sharples
    Sun Microsystems
    http://blogs.sun.com/roller/page/sharps
  8. I'm good to see Sun doing this.Am I the only one dying to know what the scores would be if the identical test was repeated, but Sun swapped out the Sun app server for Jboss, Geronimo and Caucho Resin? (Other open source servers.)

    Yes, that would be interesting wouldn't it. We actually do know what would happen because we've tried it but there's nothing to motivate us to publish anything; for one thing we'd be accused of rigging the numbers.

    If you want to see JBoss or Geronimo numbers you're going to have to ask them.

    One thing we do feel is in the public interest is the details of how we achieved the benchmark (ie. throughput and Java EE compatibity re. MySQL 5.0) - Tom Daly has already started blogging about this :

    http://blogs.sun.com/roller/page/tomdaly

    and we'll probably be producing a WebCast / Tech Article in the future.

    Rich Sharples
    Sun Microsystems
    http://blogs.sun.com/roller/page/sharps
  9. Yes, that would be interesting wouldn't it. We actually do know what would happen because we've tried it but there's nothing to motivate us to publish anything; for one thing we'd be accused of rigging the numbers.

    If you want to see JBoss or Geronimo numbers you're going to have to ask them.

    OK, be accused of rigging them. You're not the only one to have run these tests, but all you big app server vendors are afraid of getting dirt on your shoes. Just publish the numbers as a hardware vendor, like HPaq used to do.

    Either way, you can sell more servers.

    Peace,

    Cameron Purdy
    Tangosol Coherence: Clustered Shared Memory for Java
  10. OK, be accused of rigging them. You're not the only one to have run these tests, but all you big app server vendors are afraid of getting dirt on your shoes. Just publish the numbers as a hardware vendor, like HPaq used to do.Either way, you can sell more servers.Peace,Cameron PurdyTangosol Coherence: Clustered Shared Memory for Java

    We're not above doing that and we have jointly submitted benchmarks with BEA in the past; for example.

    You're assuming that the submissions would relect positively on our hardware; which raises an interesting question - I wonder why HP haven't submitted something on behalf of JBoss considering their relationship ?

    Rich Sharples
    Sun Microsystems
    http://blogs.sun.com/roller/page/sharps
  11. Cameron :
    I'm good to see Sun doing this.Am I the only one dying to know what the scores would be if the identical test was repeated, but Sun swapped out the Sun app server for Jboss, Geronimo and Caucho Resin? (Other open source servers.)

    Was this an open source app server that they tested with? IIRC, Glassfish is their only OSS app server, and the spec for it isn't even ready yet. (J2EE 5)

    What did I miss?

    -geir
  12. Was this an open source app server that they tested with? IIRC, Glassfish is their only OSS app server, and the spec for it isn't even ready yet. (J2EE 5)What did I miss?-geir

    The benchmark was done with Sun Java System Application Server
    8 Platform Edition, which is a FREE application server.

    Glassfish is Application server 9, so it shares lot of code with Application Server 8.
    Peter
    (Sun employee)
  13. It's the x86 AMD based Sun here, but they still spend R&D money on Sparc. :-(

    .V
  14. It's the x86 AMD based Sun here, but they still spend R&D money on Sparc.
    Sure - SPARC is their high end bread and butter. The AMD stuff is more for mass-market, as far as I can tell. Why would they not concentrate on their core competencies in R&D? The low-end machines don't seem to be merely a dalliance in bitty-boxen, mind you, and they are getting research dollars (see the new workstations, which are very nice indeed from what I've looked at) but...
  15. It's the x86 AMD based Sun here, but they still spend R&D money on Sparc. :-(.V

    Horses for courses. Products based on the different chip architectures are aimed at VERY different workloads and markets. Both being 64bit doesn't make them the same.

    Rich Sharples
    Sun Microsystems
    http://blogs.sun.com/roller/page/sharps
  16. Products based on the different chip architectures are aimed at VERY different workloads and markets. Both being 64bit doesn't make them the same.Rich Sharples

    From the posted STREAM benchmark for a SunFire 25K with 12 cpus, it looks each cpu in this box has a memory bandwidth of about "900".

    I believe a P4 will score about 1500 and an opteron about 2500.

    I am guessing that this means that for OLTP applications opteron is better suited.
  17. Sparc vs. Opteron bandwidth[ Go to top ]

    >From the posted STREAM benchmark for a SunFire 25K with 12 cpus, it looks each cpu in this box has a memory bandwidth of about 900.I believe a P4 will score about 1500 and an opteron about 2500.


    I agree with people that say that practice shows that much more expensive Sparc is much slower than Sun AMD boxes. Marketing says something else.

    .V

    http://roomity.com
  18. >From the posted STREAM benchmark for a SunFire 25K with 12 cpus, it looks each cpu in this box has a memory bandwidth of about 900.I believe a P4 will score about 1500 and an opteron about 2500.
    I agree with people that say that practice shows that much more expensive Sparc is much slower than Sun AMD boxes. Marketing says something else..Vhttp://roomity.com
    There are other reasons to purchase the more expensive hardware than pure performance.
  19. http://www.mysql.com/products/connector/j/

    Did you use the commercial MySQL JDBC drivers or did you use the GPL drivers, and what license obligations followed from your choice?
  20. http://www.mysql.com/products/connector/j/Did you use the commercial MySQL JDBC drivers or did you use the GPL drivers

    As always the full details of the configuration is documented with the submission :

    http://www.spec.org/jAppServer2004/results/res2005q3/jAppServer2004-20050701-00011.html#JDBCMySQL_Connector/J_3.1.90

    - the short answer is that we used the recommended MySQL Connector/J 3.1.9 driver.
     and what license obligations followed from your choice?

    Well, given that we aren't copying, distributing or modifying the drivers I don't think we have any obligations.

    Please restate your question if I'm barking up the wrong tree.

    Rich Sharples
    Sun Microsystems
    http://blogs.sun.com/roller/page/sharps
  21. First, a few disclaimers

    1) I work for MySQL AB
    2) I helped out with the JDBC side of things on this benchmark

    Okay, with that out of the way...Since Sun doesn't (and _can't_ in fact) redistribute this benchmark (because of the license SPEC places it under), there is no derived work made with the GPL MySQL drivers, and thus the terms of the GPL are not in force in this situation.

    You might want to check out http://dev.mysql.com/tech-resources/articles/mysql-network-and-you.html where we (MySQL AB) explain our position on the GPL and how it affects various scenarios, especially given our new(ish) subscription offering, "MySQL Network".

      -Mark

      (developer of Connector/J, Development Manager for Connectivity Products -- MySQL AB)
  22. First, a few disclaimers

    1) I work for MySQL AB
    2) I helped out with the JDBC side of things on this benchmark

    Okay, with that out of the way...Since Sun doesn't (and _can't_ in fact) redistribute this benchmark (because of the license SPEC places it under), there is no derived work made with the GPL MySQL drivers that is _distributed_, and thus the terms of the GPL are not in force in this situation.

    You might want to check out http://dev.mysql.com/tech-resources/articles/mysql-network-and-you.html where we (MySQL AB) explain our position on the GPL and how it affects various scenarios, especially given our new(ish) subscription offering, "MySQL Network".

      -Mark

      (developer of Connector/J, Development Manager for Connectivity Products -- MySQL AB)