Home

News: Sun Java SE 6 released

  1. Sun Java SE 6 released (32 messages)

    Sun has announced the availability of Java Standard Edition 6' final release, the result of two years of open review, external contributions, and weekly releases. Many deployers in the field, using pre-release versions, have reported noticeable runtime performance improvements, in the "two-digit range" according to Sun, which translates to anywhere from 5% to 24% improvement. Note that this is anecdotal, of course, and depends on what tasks you're performing. However, the JIT has been improved, with better runtime analysis of program characteristics, and the performance improvements have been agreed with by external users. Desktop support may not be as relevant for the enterprise community, except for rich clients. System tray support, splash screen support, and other elements have been added. JSE6 now has dynamic language support. It comes pre-delivered with Netscape's Rhino, a Javascript engine, and the scripting project's home page documents many other available scripting languages, including Ruby, Python, awk, Jelly, Pnuts, and Scheme. Java SE 6 also has new diagnostic capabilities. For example, profilers and debuggers can now attach to a running JVM without specifically using a debugging-capable configuration. For example, if a problem is found at run-time for a production server, a debugger can attach to it without restarting the server, a valuable feature for issues that show up after the server has been running for a long period. Sun is also offering sixty days of free developer support for JSE 6 through their Developer Services program. Jean Elliot, director of Developer Marketing at Sun, encouraged developers to take advantage of this offer, to expose the service to more people and to aid migration to JSE6. SDN members can subscribe to one year or three years of professional Developer Support at a 10% discount. Also important for this release, the documentation is up to date at the point of release, instead of lagging behind as Sun did with JSE 5. It's altogether a huge release. Good job, Sun. [Editor's note: this post has been updated to use the term "Java SE 6" as per Sun's request, as if none of you understood that JSE6 was the same as "Java SE 6," and to remove the note that mentions that at the time of publication the Java SE 6 home page had not been updated to reflect Java SE 6's release status.]

    Threaded Messages (32)

  2. Speed, give me speed![ Go to top ]

    We've been doing a lot of benchmarking with JDK6 for Intel on their new Quad-core Xeon. We used the 64 bit AMD Opterons (on a V40z) and the new Intel Xeons using JDK 5 and 6 (all 64 bit versions, both JRockit and Sun's JDKs) and I can tell you that Sun's JDK 6 is notably faster than 5 and it appears that Intel is way off in the lead again over the Opterons. The actual numbers will mean nothing here out of context but JDK6 on Intel is very impressive. Intel will hopefully be releasing out benchmarks very soon. -John- CTO, C24
  3. Re: Speed, give me speed![ Go to top ]

    it appears that Intel is way off in the lead again over the Opterons.
    How many cores for the Opteron?
  4. Re: Speed, give me speed![ Go to top ]

    How many cores for the Opteron?
    This was 4 cores vs 4 cores, but of course the clock speeds and bus speeds weren't identical. Given the fudge-factors and various other fizziness the Intel's were still quite a bit faster (per core per GHz). -John-
  5. Re: Speed, give me speed![ Go to top ]

    How many cores for the Opteron?


    This was 4 cores vs 4 cores, but of course the clock speeds and bus speeds weren't identical. Given the fudge-factors and various other fizziness the Intel's were still quite a bit faster (per core per GHz).

    -John-
    Well, clock speeds are pretty much BS anyway. I would think the best way to do this would be per dollar and/or per watt.
  6. Re: Speed, give me speed![ Go to top ]

    Well, clock speeds are pretty much BS anyway. I would think the best way to do this would be per dollar and/or per watt.
    This is exactly why we were asked to do it for Intel. Our FpML, FIX, ISO-20022 and SWIFT performance figures were quoted at the Quad-core launch at the top of the Gherkin in London the other week. Of course what the banks are interested in is the performance per $ (or £ in the UK), not the cost of the box but how much heat they have to get rid of. Obviously the C24 computer room is not designed to measure these sorts of things but we were able to benchmark raw throughput on the boxes we were given. I'll post some basic figures either later tonight or tomorrow but they will not be quaified, Sun's figures of "around 24%" are in line with what we've found from 5 to 6. -John-
  7. Re: Speed, give me speed![ Go to top ]

    We've been doing a lot of benchmarking with JDK6 for Intel on their new Quad-core Xeon. We used the 64 bit AMD Opterons (on a V40z) and the new Intel Xeons using JDK 5 and 6 (all 64 bit versions, both JRockit and Sun's JDKs) and I can tell you that Sun's JDK 6 is notably faster than 5 and it appears that Intel is way off in the lead again over the Opterons...
    Any benchmark with specs? Andy
  8. Re: Speed, give me speed![ Go to top ]

    Any benchmark with specs?

    Andy
    This will be published with source code and machine specs but there are rules about "official" benchmarks and we have to fully qualify everything, it won't be long now. One of the interesting problems we're having it the JIT, how do you warm the JIT up and know if it's fully optimised? We're also working with BEA to tune JRockit, lots of stuff to do and as usual never enough time. -John-
  9. Re: Speed, give me speed![ Go to top ]

    I discovered the Javolution website earlier today: javolution.org. The website claims that the Javolution project provides fast implementations of java.util.List and java.util.Map. Is it pure marketing talk? Or should we all switch our HashMap to FastMap?
  10. Re: Speed, give me speed![ Go to top ]

    "The website claims that the Javolution project provides fast implementations of java.util.List and java.util.Map. Is it pure marketing talk?" No, those implementations are faster. The trick they use is simple: reuse objects. This is because the creation of new objects in Java is very expensive.
  11. Re: Speed, give me speed![ Go to top ]

    "The website claims that the Javolution project provides fast implementations of java.util.List and java.util.Map. Is it pure marketing talk?"

    No, those implementations are faster. The trick they use is simple: reuse objects. This is because the creation of new objects in Java is very expensive.
    What objects are you talking about, internal immutable objects using flyweight? Or are they doing some caching of objects that are stored? In that case, it's only faster in certain use case instances. Ilya
  12. nope[ Go to top ]

    creation of new objects are not that very expensive in Java, especially comparing to C++.
  13. Re: nope[ Go to top ]

    creation of new objects are not that very expensive in Java, especially comparing to C++.
    That depends on a lot of factors. It can become very expensive in environments where thousands and/or millions of value objects are created. Though the reason to try to make your object immutable or at least limit mutability. In that case, you can use flyweight and other creational patterns.
  14. Re: Speed, give me speed![ Go to top ]

    I discovered the Javolution website earlier today: javolution.org.

    The website claims that the Javolution project provides fast implementations of java.util.List and java.util.Map.

    Is it pure marketing talk?
    Or should we all switch our HashMap to FastMap?
    I guess those implementations are faster - on a certain micro benchmark level. But I'm sure, that someone could create micro benchmarks where they are not faster. The main problem with benchmarking such implementations is, that they behave very differently on different platforms and scenarios in which they are performed. Further, I _strongly_ doubt, that you gain significant performance improvements in client-server applications by using a whatsoever fast collection implemention. In fact over 95% of the performance in typical C/S applications is lost in two places: network/remoting communication and database access. So, no, we should not all switch to FastXYZ (at least I will not). Regards, Dirk PS: Javolution also provides means for XML binding, or at least for XML-Object (de-)serialization. I would strongly recommend not to use this, since it is buggy as hell (speaking of version 4.0). PPS: Before using _any_ open source library, one should check how big the user community is. In case of javolution, for me this is an absolute kick-out criteria.
  15. Re: Speed, give me speed![ Go to top ]

    As far as I know Javolution is an API for real time in Java. So they have implemented some collections that were faster than the Sun ones with preceding version of the jdk. I don't know if there is still a gap. You'd better test it by your own. I know by example that some code snippets that were faster with jdk 1.2 or 1.3 (like implementing an object pool) are not effective with the last java versions. I had a look on their website and they claim they do it. In any case it is a best practice to test the impact on your application. Best regards, Bruno CHEVALIER
  16. Re: Speed, give me speed![ Go to top ]

    I think you forgot the disclaimer (who founded your benchmarks, they purpose and destination, etc...). Also, Intel Xeons will always be much faster than Opterons when running on JRockit since Intel is providing BEA with heaps of cash in development funds. That also explains why the when to develop the only JVM around for Itanium and why the JRockit guys are always touting their performance of Xeons and Itaniums. I would think that for 1 and 2 CPUs and low memory usage Intel should be faster. But I would be surprised if for 4 CPUs and up or for applications using memory heavily. Opterons didn't come on top. But this is only based on each CPU's architecture and theory. I haven't perform any benches. I'm waiting for your numbers and a disclaimer for a more informed opinion. My own disclaimer. I own AMD stock, and I don't work or ever worked for Intel, Sun, AMD or BEA.
  17. Re: Speed, give me speed![ Go to top ]

    Intel Xeons will always be much faster than Opterons when running on JRockit since Intel is providing BEA with heaps of cash in development funds.
    You could use a similar argument for Sun and SPARC/AMD chips. Why would Sun give a dime about optimizing Java for hardware they don't sell? Note that I don't know if Sun reasons this way, but it would make sense economically.
    That also explains ... why the JRockit guys are always touting their performance of Xeons and Itaniums.
    And AMD :-) Our latest speed bump gave a 30% performance bump on SPECjbb2005 on Opteron (which is on par with the increase on Xeon). And more on more "micro" benchmarks. I will cover JRockit/Opteron performance in my blog as soon as I can spare the time.
    I would think that for 1 and 2 CPUs and low memory usage Intel should be faster. But I would be surprised if for 4 CPUs and up or for applications using memory heavily. Opterons didn't come on top. But this is only based on each CPU's architecture and theory. I haven't perform any benches. I'm waiting for your numbers and a disclaimer for a more informed opinion.
    SPECjbb2005 is an extremely memory intensive benchmark, go check the scores here. As is SPECjAppServer2004 with results here. Of course, these are only a couple of proof points, you will want to check other benchmarks as well. And my disclaimer: I work in the BEA JRockit team... Cheers, Henrik
  18. Re: Speed, give me speed![ Go to top ]

    what about Niagara? I had the pleasure of speaking with Dr John Busch who heads the CAD team at Sun (incredibly interesting guy, and probably one of the best presentations I have heard).
  19. Re: Speed, give me speed![ Go to top ]

    My own disclaimer. I own AMD stock, and I don't work or ever worked for Intel, Sun, AMD or BEA.
    Well I only own stock in my own company but as with all stock it's always best if you hold on to it for the long term. If Intel is faster than AMD today (and I believe it currently is) they'll take the lead again in the near future. I think you'll find that Sun also did a lot of work with AMD to optimise the JVM for the Opteron. C24 are partners with Sun, Intel and BEA so I have to be careful what I say. The benchmarking wasn't funded by any of the above. They are ours for our own marketing use and for that reason we've used Opteron, Intel, Sun's JVM and BEA's JVM. We simply collated some figures together for Intel's Quad-Core launch the other week and I mention it here because they were very impressive, the fastest I've seen to date. -John-
  20. Re: Sun JSE 6 released[ Go to top ]

    And I look forward to getting it on my Mac sometime in 2008.... geir
  21. Re: Sun JSE 6 released[ Go to top ]

    And I look forward to getting it on my Mac sometime in 2008....

    geir
    Hi Geir, That's probably about the biggest reason I've yet to make the move to Mac, I'll get there one day, 2008 you say? -John-
  22. Re: Sun JSE 6 released[ Go to top ]

    And I look forward to getting it on my Mac sometime in 2008....

    geir


    Hi Geir,
    That's probably about the biggest reason I've yet to make the move to Mac, I'll get there one day, 2008 you say?

    -John-
    Bunch of FUD. Most people hesitate, but when they switch, they dread the day they ever have to look at windows again.
  23. Re: Sun JSE 6 released[ Go to top ]

    <blockquoteBunch of FUD. Most people hesitate, but when they switch, they dread the day they ever have to look at windows again.</blockquote> I agree, I'm mostly Linux and have been for years, I'll do it one of these days. Wrong thread to go into Mac hardware vs. other but I'd be interested to hear from anyone that can give me some benchmarks of Java 5 or 6 on Mac vs. the same on Windows or Linux. I'm happy to provide some code if anyone's got a machine with the JDKs and 15 minutes to spare. -John-
  24. Re: Sun JSE 6 released[ Go to top ]



    I agree, I'm mostly Linux and have been for years, I'll do it one of these days. Wrong thread to go into Mac hardware vs. other but I'd be interested to hear from anyone that can give me some benchmarks of Java 5 or 6 on Mac vs. the same on Windows or Linux.

    I'm happy to provide some code if anyone's got a machine with the JDKs and 15 minutes to spare.

    -John- John, I'd be happy to test on my macbook pro. Send me the code. You can get my contact info from http://www.ilyasterin.com/about.html Ilya
  25. Re: Sun JSE 6 released[ Go to top ]

    John, I'd be happy to test on my macbook pro. Send me the code. You can get my contact info from http://www.ilyasterin.com/about.html

    Ilya
    Thanks you have mail, watch this space folks :-) -John-
  26. Re: Sun JSE 6 released[ Go to top ]

    And I look forward to getting it on my Mac sometime in 2008....

    geir


    Hi Geir,
    That's probably about the biggest reason I've yet to make the move to Mac, I'll get there one day, 2008 you say?

    -John-


    Bunch of FUD. Most people hesitate, but when they switch, they dread the day they ever have to look at windows again.
    Absolutely. I've been on OS X since the first ibook had it, and never looked back. Ok, I started to carry a winXP thinkpad around for a while for corporate email, but still carried my powerbook or macbook too... geir
  27. Re: Sun JSE 6 released[ Go to top ]

    And I look forward to getting it on my Mac sometime in 2008....

    geir
    I don't know what OS X you have, but I've had JSE 6 on my Tiger for months now. The default is still 5.0, though of course you can manually change that. I'm sure they'll release the final 6.0 in the coming weeks, but you can download it from their adc site. Ilya
  28. Re: Sun JSE 6 released[ Go to top ]

    And I look forward to getting it on my Mac sometime in 2008....

    geir


    I don't know what OS X you have, but I've had JSE 6 on my Tiger for months now.

    The default is still 5.0, though of course you can manually change that. I'm sure they'll release the final 6.0 in the coming weeks, but you can download it from their adc site.

    Ilya
    I didn't realize it was on ADC. Clearly 2008 was an exaggeration... Still, it felt that it took forever to get JDK5 GA on OS X... geir
  29. Performance delivered![ Go to top ]

    I quickly put new JavaSE 6 under simple distributed computing benchmark (includes RMI/CORBA/WebServices/ICE) and the results are excellent. The performance improvement varies from 5-50% with the average at about 15-20%. All tests were on the same machine using server VM, for both JavaSE 5 & 6.
  30. Java Generics & Collections[ Go to top ]

    If you're interested in some of the new features from 5 and 6 then try the new book from Maurice Naftalin "Java Generics and Collections", it's an excellent book. -John-
  31. Re: Java Generics & Collections[ Go to top ]

    If you're interested in some of the new features from 5 and 6 then try the new book from Maurice Naftalin "Java Generics and Collections", it's an excellent book.

    -John-
    Yes, excellent book from random chapter readings. I'm yet to read it from cover to cover, still have a few other books to finish before, but I look forward to doing so in the next few weeks. Ilya
  32. Java Generics & Collections[ Go to top ]

    If you're interested in some of the new features from 5 and 6 then try the new book from Maurice Naftalin "Java Generics and Collections", it's an excellent book.

    -John-
    Well, by Maurice Naftalin and Phil Wadler. We do have some coverage of the Java 6 collections. It's a surprisingly active area. ArrayDeque is very fast; they think it's set to become the new mainstream Collection implementation, though IMO it was a mistake not to have made it implement List. Check out NavigableMap and NavigableSet too; they extend (and obsolete) SortedMap and SortedSet, with a lot of extra useful function. They've not stood still for this release! Maurice
  33. Source release..[ Go to top ]

    It seems source release is not updated(I checked here http://download.java.net/jdk6/).