Discussions

News: New ECPerf Results: Pramati App. Server 3.0 Results Posted

  1. New ECPerf Results for Pramati have been posted to TheServerSide. Pramati Application Server 3.0 SP1 was run on a Compaq DL-580 using Oracle 8.1.6 to yield a respectable Price/Performance figure of $14 Price/BBops and a Performance figure of 4018.0 BBops/min@Std.

    Read Pramati's ECPerf Announcement

    Check out Pramati's ECPerf Results
  2. As I read in the Pramati announcement, the App. Server takes only 20% from the total 14$/BBops. Can anybody explain me, where the other 80% come from. I think the Intel based server should be cheaper than the Unix boxes and BEA used Oracle, too. So what costs fill the gap?

    Mirko
  3. There is a cost split up graph which is shown in the " ECPerf FAQ" in Pramati's ECPerf announcement.

    http://www.pramati.com/docstore/1300002/ecperf_faq.htm
  4. <snip>
    As I read in the Pramati announcement, the App. Server takes only 20% from the total 14$/BBops. Can anybody explain me, where the other 80% come from. I think the Intel based server should be cheaper than the Unix boxes and BEA used Oracle, too. So what costs fill the gap?
    </snip>
    The primary difference comes in the power of the hardware. The new (P4)Xeon is much more powerful (3 times the speed, and a much higher-performing architecture) than the 700MHz PIII Xeon. This with a bigger bonus of coming at a much lower price (Moore's law at work). Added to that, Oracle license is per CPU. So on newer Xeon, u can do more with lesser number of Oracle licenses.

    700MHz Xeon was chosen for the first submission here for strategic reasons (customers use existing hardware for their initial app deployments. 700MHz Xeon based servers is widely available in organisations now).

    Cheers

  5. Congratulations to Pramati for posting results comparable to the big boys!

    Where are oracle, HP, and IPlanet?
  6. Bravo,
    It's great that they published a number.

    On a sidebar
    What becoming pretty interesting is that no one seems keen to use a Sun JDK for any of these runs. Pramati used IBM's JDK for their run. While I understand why, this has to be worrying Sun that J2EE vendors doen't seem keen to do runs on their hardware with a Sun JDK. Maybe, when 1.4 ships any issues will be resolved and we'll see some runs.

    Billy (works for IBM)
  7. --- START QUOTE
    On a sidebar
    What becoming pretty interesting is that no one seems keen to use a Sun JDK for any of these runs. Pramati used IBM's JDK for their run. While I understand why, this has to be worrying Sun that J2EE vendors doen't seem keen to do runs on their hardware with a Sun JDK. Maybe, when 1.4 ships any issues will be resolved and we'll see some runs.
    --- END QUOTE

    This is probably not the case, but the thought crossed my mind when I read that. Maybe Sun's long term goal is to get out of the VM business. Why not just license out, and manage the evolution of the spec, without an implementation (or at least a good one) in house. It may be more profitable to be the IP (as in property) holder, while doing none of the major development work. Since Sun gives away the JVM for free, it's a loss leader anyway.

    It's like Microsoft still getting all the royalties for products sold, but not spending the man-hours on product development of an end user product.

    The way Sun makes money off of java has always puzzeled me since it started.

    -Pete