Discussions

News: BEA announces JRockit 8.0: Supporting Intel Itanium 2

  1. BEA has released an upgrade to its JRockit JVM for servers based on Intel processors. The main enhancements in the new release of JRockit 8.0, are in performance, and the profiling and debugging interfaces to help developers write faster apps and to find and fix performance bugs.

    BEA WebLogic JRockit 8.0 is the first release of a standalone JVM optimized for the Intel Itanium 2 processor. In addition to enhanced support for the IA-32 architecture and further optimization for the BEA WebLogic Enterprise Platform.

    There was also an interesting comment about the growing interest in WebLogic on Intel:

    "Most of BEA's customers run its WebLogic Java application server on Unix systems from the likes of Sun Microsystems and Hewlett-Packard, but sales on Intel-based systems represent the fastest growing part of its business, said Bob Griswold, vice president and general manager of BEA's Java run-time group"

    Read the JRockit 8.0 release article

    Read the BEA Press Release

    Have many people played with JRockit?
    Did it manage to get you to switch from Suns or IBMs JDK? If so, why?

    Threaded Messages (21)

  2. wow! support for intel?[ Go to top ]

    The press release was kind of funny, since jRockit has only really been available for Intel platforms (although internally they supposedly did work to support Sparc etc. as well.)

    I've seen the new jRockit in action (I thought it was a beta of 8.1 but I may be confused) and while it works pretty well, the coolest stuff is what is outside the JVM ... there is a complete management console now that can be run separately to manage any number of JVMs. So, for example, you can have it beep you when an out-of-memory exception happens, etc. You can monitor all the stats kind of like in the Windows Task Manager (the graphical "top".)

    Last year at eWorld, the BEA "inside talk" on the jRockit purchase was that they needed a very stable JVM to run on Intel (Windows and Linux both, I assume,) which seems to imply that they didn't feel that the existing JVMs for x86 met that criteria. So while the jRockit stuff may be fast (in our tests, it is sometimes faster, sometimes not,) I understand that the main goal of jRockit is to work reliably and be an acceptable part of the "enterprise infrastructure" (hence features like the management console.)

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!
  3. JRockit nice with WLS 7.0[ Go to top ]

    I have used JRockit 7.0 as the JVM for WLS 7.0 for a benchmark. Compared with the JVM from Sun, it turned out to be faster and more robust, in the sense that it can support many more execute threads without crashing (Sun's JVM litterally froze after 50, repeatable experiment).

    JRockit definitely seems to be the JVM for Windows/Intel platforms. I'll try with 8.1 one of these days to see if I can spot a difference in performance.

                    Yann
  4. The mgmt console stuff is neat, but I am going to hold off on upgrading unless I can find out what, if any, performance improvement to expect. We are supporting a computationally intensive rule engine and have settled on JRockit as our VM, based on a minor amount of benchmarking against other VM's.

    The release notes don't mention anything about performance enhancements, new tuning mechanisms or anything...
  5. What are the performance enhancements?[ Go to top ]

    Has anyone tried to get JRocket running with WebSphere 4.x. I know people have had some success with the Sun Windows JVM. Just curious and getting frustrated with Big Blue's lack of urgency in updating their JVM.
  6. Griswold[ Go to top ]

    Ok. So Girswold says:

    "It's the world's fastest JVM. We've made that claim before, but now it really is," Griswold said.

    Anyone else see the problem here? Does this mean that up until now its all been a lie?????
  7. Following the steps of Webgain?[ Go to top ]

    Reminds me of the time when WebGain was the official IDE for Weblogic. Always 6+ month behind, debugging would work only with their special JDK, buggy ...

    Where's WebGain now? Where's support for JDK1.4 in Weblogic? Remote debugging between IDE on JDK1.4 and WLS on JRockit is working fine, ah?

    Anyway, we are more interested in JBoss than in JRockit.
  8. Following the steps of Webgain?[ Go to top ]

    Last month Fortune magazine have a detailed report about 64 bit computer processing and the status of company CEOs who decided to follow Intel's Itanium project. Seems like Intel Itanium 1 was a big flop. Intel has given an assurance that Itanium 2 will solve all the outstanding issues. But there are number of firms such as AMD and IBM which never followed Intel's dream world calculations 64 bit computing and decided to follow their own 64 bit computer design and research. HP's CEO decided to stop all processor related research work and wait for Intel's dream processor. I think even BEA also decided to follow the same path cuz they don't have any other option. I am not much worried why BEA decided to follow HP, because BEA is a company living with acquiring of small and cheap products available in this world. Anybody who look into the history of all major products from BEA can easily know that they have acquired them from others firms.

         Intel Itanium 2 seems like good only in paper and dreams. In performance and price AMD and IBM and even sun processors have better future. Fortune has clearly mentioned that AMD’s new processors are far better and they are only bad in marketing their ideas. Recently Dell agreed to pack AMD’s 64 bit processor in their new servers but vendors such as HP have to wait for Intel’s dream processor which failed completely to meet customer’s expectation in its first version (Itanium 1). I think Intel and their partners in Itanium project use BEA as marketing and publicity window. Until and unless Itanium 2 processor is proved as a successful industry standard there is no use of BEA’s large-scale marketing on their preparatory JVM. As its name indicates, it might be a 20 years old rocket technology colored with some new Graphic user interface to sell among American corporate managements.

         Finally this is my personal opinion and I don’t want people who do marketing job for BEA to start another Internet fight with me on this subject.
  9. Personal Opinion[ Go to top ]

    "because BEA is a company living with acquiring of small and cheap products available in this world"

    They were clever enough to buy Tuxedo.
  10. Re: Personal Opinion[ Go to top ]

    "because BEA is a company living with acquiring of small and cheap products available in this world"

    >
    > They were clever enough to buy Tuxedo.

    And they were clever enough to buy WebLogic.
  11. Following the steps of Webgain?[ Go to top ]

    'Official' WLS platform JDK1.4 support has been a thorn in my side as well - though I'm only aware of a few relatively minor side-effects when running Server 7 on 1.4. Upcoming 8.1 platform release is supposed to be JDK1.4 based. It sounds like the 8.1 beta will be available in a few weeks.

    I'd love to know the inside story on why JRocket for Sparc is not available - though it is rumoured to exist. Political (as opposed to technical)reasons are my guess... ah well - quit dreaming and back to coding!

    cheers,
    Markus
  12. What are your experiences with JVM's and linux? I will probably use Resin with Red Hat linux soon and have been using Sun JVM during development.

    How does JRockit compare to the Sun JVM, in terms of performance and scalability? Is it true that the Sun JVM has issues after fairly low connection counts?
  13. World's fastest JVM?[ Go to top ]

    BEA have been claiming for a while that JRockit is the "world's fastest server-side JVM" but I've been struggling to find any publicly available benchmarks which illustrate this.

    You'd have thought BEA would want to rub Sun's nose in it with some groovy graphs and some top ten lists but there don't seem to be any on the BEA site. This makes me a bit dubious.

    Has anyone found anything definitive?

    Just how much faster is JRockit compared to Hotspot?


    Robbie
  14. jRockit Benchmarks[ Go to top ]

    See the SPEC site for JVM benchmarks. The jRockit numbers are very impressive.

    http://www.spec.org/jbb2000/results/jbb2000.html

    Make sure to make meaningful comparisons i.e. same number of CPUs, etc. All in all it looks like BEA has a strong case for thier performance claims.

    And, as Cameron said, the monitoring aspect is really cool.

    All in all if they can beat the others in performance and administration, it sounds like a winner to me.

    Harvey
  15. jRockit Benchmarks[ Go to top ]

    ... or at least it gives some thought to the "other guys" who went to sleep at the switch ...

    (It's time Sun and IBM put their JVM programs back in gear. Or if they are in gear, they need to make some more noise ;-)
  16. jbb2000 results[ Go to top ]

    Harvey -

    Fair enough - I only glanced at the jbb2000 results yesterday and couldn't really make much sense of them. However, after your suggestion I went back and did the math.

    It seems that JRockit 8 is 20% faster than Hotspot (in a 4 CPU configuration) which is pretty impressive (especially as it looks like the Hotspot environment was running with twice as much memory!). However, it's only 4% faster than the 4-CPU IBM JVM which isn't quite as impressive...

    My main issue with these results though is that since each JVM was tested using different hardware it's virtually impossible to say definitively what the fastest JVM is as there's no way of knowing how much the performance of the hardware itself contributes to the overall result.

    If I'm going to buy JRockit I want to know that the apparent 20% improvement in performance I'm buying is actually going to apply in my environment - I don't want to have to buy all the hardware too!

    And on that subject, my next question is how much is JRockit? Pricing was another crucial bit of information which seemed to be missing from the BEA website when I looked yesterday.


    Robbie
  17. jbb2000 results[ Go to top ]

    From the previous announcements, I thought that jRockit was no charge, and haven't seen an indication otherwise, however I would not assume it ;-)

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!
  18. jbb2000 results[ Go to top ]

    JRockit is free.
  19. jbb2000 results[ Go to top ]

    I looked at the SPEC.org results as well. It looks like the closest SUN performance on a 4-way server is still 50% off the recent JRockit results. Quite impressive. I understand where you are coming from Harvey, but with Intel and JRockit, you seem to get more choice since you are not tied to a specific hardware platform. You can choose among low-cost hardware solutions and achieve these great performance results.
  20. jbb2000 results[ Go to top ]

    I looked at the SPEC.org results as well.


    Really? Then you should've seen this:

     specjbb2000, 4 processors
     76136 Weblogic JRockit 8.0 on Intel Xeon MP 2000 Mhz
     73319 J2RE 1.4.0 IBM Windows 32 on Intel Xeon MP 2000 Mhz
     63414 Hotspot 1.3.1.08 on HP-UX 11i v1.6 for Itanium 1000 Mhz

    And that's Hotspot 1.3.1(!) on 1000 Mhz(!) Itanim(!) (see Intel Itanium 67% slower than Alpha in SAP tests)

    It appears very unlikely that JRockit would outperform Suns Hotspot 1.4.1 if the tests were done on the same hardware.
  21. Which JVM scales the best before crashing?[ Go to top ]

    Which JVM can handle the most threads/connections on a Linux system before choking?

    Sun, IBM, JRockit, Black(somthing)?
  22. it's old, but ...[ Go to top ]

    Osvaldo's last report that covered Linux is from 2000:

    http://www.javalobby.org/members/jpr/part2a.jsp

    I think that you can download Volano's benchmark (lots of threads and sockets) and test as well.

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!