Sun, JBoss Go Back to Butting Heads Over Java?

Discussions

News: Sun, JBoss Go Back to Butting Heads Over Java?

  1. Sun, JBoss Go Back to Butting Heads Over Java? (34 messages)

    It seemed that Sun and JBoss Group may have reconciled their differences back in July. However, this latest report claims that there is still be a rift. Jonathon Swartz stated, "There are some duplicitous companies that like to compare and equate open source with not-for-profit."

    This is an obvious jibe at the fact that although JBoss the application server is open source, JBoss Corp is behind it making money, and Sun isn't a fan.

    On the other side, Marc Fleury said that Sun appears to be two-faced in the way it doles out J2EE licenses. He cited the JOnAS/ObjectWeb effort, and how it received the J2EE license for free, so he believes his company is entitled to a free or low-cost licensing deal. He also noted that while JOnAS is a not-for-profit entity, it has the backing of several companies—including Groupe Bull and France Telecom—that are for-profit.

    Read Sun, JBoss Continue to Butt Heads Over Java

    Read the old thread: JBoss issues statement on J2EE Certification: "fully committed"

    Threaded Messages (34)

  2. Compatibility Scholarship Program?[ Go to top ]

    Did JBoss apply for the TCK via Sun's Compatibility Scholarship Program?

    http://java.sun.com/scholarship/

    {{

    Numerous implementations of Java technology Specifications are developed by not-for-profit organizations or by efforts at universities and colleges around the world. In recognizing the value of these efforts Sun is establishing the Compatibility Testing Scholarship Program to ensure that the Technology Compatibility Kits (TCK) for Sun's Java technology specifications are available and accessible to these efforts - including support services.

    The Compatibility Testing Scholarship Program is an annual support scholarship program to provide suitably qualified efforts access to Sun's TCK (Technology Compatibility Kit) support services for Java technology Specifications free of charge. A Review Board which oversees the Program determines which efforts are eligible for a support scholarship. The program will be funded at or above a rate of 30 efforts per year (current value of approximately $1M) for at least the next 3 years. The Review Board may also grant TCK access at no cost without support services to qualified efforts.

    [...]

    Both organizational and individual efforts can apply for a scholarship. Successful applicants must meet certain legal qualifications and the effort the applicant is undertaking must meet certain criteria. The legal requirements are:

    A legally organized not-for-profit (such as a '501 [c][3]' corporation) that is not owned or effectively controlled, based on prevailing standards of law in the applicable jurisdiction, by a commercial interest, that uses the TCK for the purpose of developing and distributing a compatible, non-commercial, independent implementation of a Spec.

    An individual not acting for or on behalf of a legally organized entity that uses the TCK for the purpose of developing and distributing a compatible, non-commercial, independent implementation of a Spec.

    Non-commercial means an implementation that is not used by the applicant for direct or indirect commercial or strategic gain or advantage.

    }}
  3. Compatibility Scholarship Program?[ Go to top ]

    Did JBoss apply for the TCK via Sun's Compatibility Scholarship Program?


    I don't think they are able to (read below).

    > Non-commercial means an implementation that is not used by the applicant for direct or indirect commercial or strategic gain or advantage.

    If I'm not mistaking the JBoss Group is in it for the profit, in the sense of strategic gain and advantage. Also is the JBoss AS an indirect commercial gain for the JBoss Group since their business plan is based on consultancy and trainings AFAIK.

    It could be that I'm wrong, but I can't imagine that this posiblity has allready been reviewed by the JBoss Group.

    Just my 2ct.
  4. Article[ Go to top ]

    Sun has a problem with jBoss for some problem it seems and are out to get them. I suspect Sun got ASF Geronimo to counter jBoss. jBoss has had some fine contribution to the community.
    Sun should not be viwed as a good citizen of the Java Community, but as a MS-want-to-be.

    I see open source as "for-profit" and commercial as anti profit. Some companies have 40 concurrent users and only a few servers, they have Oracle for $45 K, BEA for $45K, and run it on slow Sun boxes, BluePrints/EJB and use MS Office for $200 per at stations, and Active directory server for $15K, Excahgne for $8K etc. This is some corpporate department that is on the cost side of the company chart of accounts. So there are companies still not out of business that have a "you get what you pay for" thinking.

    Others use Linux, PostGreSQL, OpenOffice, Struts, Tomcat, iBatis DAO and faster "x86" boxes. They have racks of servers all for minimal cost. The revenue they generate from web apps via advertising, or savings or sales ends up on the revenue side of chart of accounts.

    .V

    repost:
    Who is behind SCO?
    http://news.zdnet.co.uk/software/applications/0,39020384,2137395,00.htm

    What this means?
    http://newsforge.com/newsforge/03/09/09/2355214.shtml?tid=11

    And finaly:
    http://finance.yahoo.com/q/bc?s=SUNW&t=2y&l=off&z=m&q=l&c=
    They keep losing market share in HW and iPlanet is no where, another reason to avoid them from Community; no need to hold them up.
    Looking at TPC.org, they are expensive and slow, relative to Linux on x86.

    Anyway, SUN JVM I avoid, and use J:Rockit.
  5. Article[ Go to top ]

    I find this a very limited point of view.
    I agree that Sun ain't making the JBoss Group's life an easy one, but you are making it a religious war. In the years i've worked in the corporate IT sector I've learned one thing; no matter how much everyone in the company support OpenSource (including the CEO, making it mandatory to support OS projects a day a week), it always comes down to 1 thing: we're all in it for the money and the pleasure. Sun, MS, JBoss Group, they're all in it for the money, no matter in what way and can you blame them?

    p.s. Last time i checked those Sun enterprise servers outperformed those x86 multi-proc xeon's, but yes, you'll have to pay for it.

    p.s. this is NOT meant as a flame, just a way to soften the discussion that has been held alot of times and will go nowhere in the end.
  6. Article[ Go to top ]

    Barre,

    Yes, it is shocking that more expensive is slower, but that is Sun.
    There are many tests on TPC.org, ex:
    http://tpc.org/tpcw/results/tpcw_perf_results.asp

    Even TPC C show that "x86" Linux machines are much faster and much cheaper than Sun boxes.

    Please point out to me what test shows that an expensive Sun machine is faster than a cheaper x86 ( decent such as http://www.pc.ibm.com/us/eserver/opteron
     or NewISys 2100) box. (Even a Dell box beats it on performance. Some laptops I think are faster than Sparc boxes runing SlowLaris)

    I also make clients aware about Sun/SCO relation.

    You are mistaken if you think companies pay to "work one day a week on open source", that is not how open source gets used. People just use open source becuase it is more reliable and better supported, and it has a profitable license. Profit is good. People just USE open source. CTO does not say use open sofware. Good software engineers just find a way to build less expensive/more reliable, just like a car engineers find a way to make a car more fuel efficient, weigh less, less wind resistance and costs less… . CTO just sees cost savings and revenu generated.

    See:
    http://news.netcraft.com/archives/web_server_survey.html
    Why pay MS $15K for active directory? ( and yes, Sun is listed).
    There is safety in numbers, who knows when a comerical vendor goes out of business or changes the business model.


    .V
  7. Article[ Go to top ]

    Vic,

    it's not the case that I am pro-sun or pro-JBoss, I'm pro what my client needs.
    The hardware, could be that I'm wrong about that, my mistake (and since when is the fastest the most expensive?).

    My point was that I do not agree with the "company foo is a commercial company so it sucks and we should ban them from our list of friendly companies" remarks that have been circulating the open-source community for way too long.
    I just don't get it that companies are being bashed for being commercial, if their stuff doesn't meet the standards, let's take the all-time-favorite example microsoft, I in turn say the company is on my "not-so-nice-company-list" ;-), but not for being commercial, I say it based on their aggressiveness in marketing and their inferior products / code. That's the nuance I'm looking for in the discussions. And, again, I'm not saying Sun does not make (big) mistakes or that they shouldn't give the oppertunity to JBoss to license it in a "normal" way or what so ever.

    Besides I agree that (good) software engineers find ways to make things more efficient (cost efficient, performance efficient, more maintainable, etc.) and wether that is by use of open-source components, commercial components or own hacking it's all the same to me.

    p.s. Yes, I have seen up close and personal a CEO saying that we (a small group of developpers) are allowed to work on open-source projects once a week, it was part of the benefits package so to speak, and yes we got paid during those hours. (this is just an example I've seen once, never seen it at other jobs).

    p.p.s. Yes, I am pro open source, but on grounds of my own reasons (openness, extendibility, the spreading of knowledge, the OS principle that I like, etc.), and a great deal of the time of the "bigger" OS projects the quality is decent to good, and if not, change it and submit it back. ;-)

    With kind regards,

    Barre
  8. Article[ Go to top ]

    p.s. Yes, I have seen up close and personal a CEO saying that we (a small group of developpers) are allowed to work on open-source projects once a week, it was part of the benefits package so to speak, and yes we got paid during those hours. (this is just an example I've seen once, never seen it at other jobs).

    >

    Well as the CEO of JBoss Group that is exactly our model, only I allow the developers to work FULL-TIME on Open Source, not just one day a week. That is what I call PROFESSIONAL OPEN SOURCE as opposed to hobbyist open source.

    I recently recruited REMY MAUCHERAT of Tomcat 5 fame. Also GAVIN KING of hibernate fame was recruited recently.

    More to come, we believe in giving the opportunity to leading open sourcers to become pro as employees of JBoss Group, an opportunity I never had growing up but I can offer to the best in their field today. That's the future, I am proud of that.

    marcf
  9. What is JBoss anyway ?[ Go to top ]

    Maybe Marc you don't understand very well what "getting the certification" means.

    Suppose Objectweb would get the scholarship and Jonas would be certified.
    ObjectWeb is a non-profit foundation. It's open to everyone. Jonas is LGPL. So what ?
    Even JBoss could do business around Jonas. Even JBoss could join Objectweb.
    What's inconsistent in that ? Certification applies to a piece of software, not
    to a company. Scholarship applies to an organization, not to lines of code.

    The thing is, one never knows if "JBoss" refers to a company or a project. Do the ads up this page apply to JBoss-the-server or JBoss-the-company ?
    Sounds like a clever little for-profit guy set up a company with a front open-source-but-not-open-minded project :)

    Suppose Geronimo in some remote future would get the certif too, Apache
    getting the scholarship... The JBoss Group could just do the same and make some money from Geronimo.

    Red Hat did not join Objectweb to get Jonas : it's already open-source ! Just use your brain :-)
    They put pretty clearly that they joined to donate some code and contribute to the development effort.
    Objectweb was their only choice : the only application-server to be fully operational and to be developped
    in a fully open way.

    By the way, you should notice that your old friend the Apache foundation is packed with staff from Big Blue, RedHat, Collabnet and more profit-making companies - and is not even as open as Objectweb is, since it's a "meritocracy" ie a gang of buddies. Why don't you get mad at this Geronimo goblin that braggs about getting certified one day, cover his ass by setting up this scholarship trick with Sun a year ago and started up stealing your own code base :-) ?

    Now let's imagine JBoss (the company, not the eponymous project) would get the scholarship.
    Could anyone (Red Hat and al) get aboard JBoss (the project, this time) ? Well,
    according to recent events, everybody doubts it. Maybe this is the real difference between JBoss' status and Objectweb's.
    Objectweb is an open non-profit whereas JBoss is not (neither the company nor the project).

    There's nothing to get mad about.

    As for me, I wouldn't offer to pay, since they are vibrant non-profit projects around, that may be certified soon, without anybody spending mucho dolares - and without you making your pile with my bucks and sweat.
  10. My dubious statistics[ Go to top ]

    The committer list for Geronimo stands at 15 people (yes I know this is out of date but I can't find a newer one). Of this 15, 8 work full-time for Core Developers Network, and one more in an associate capacity.

    The committer list for JBoss consists of 85 people, of whom roughly 30 (from Marc Fleury's quotes here - jboss.org is down right now so I can't get a full list) work for JBoss Group.

    In other words, (assuming a constant commit-bandwidth per developer!), a single commercial entity (CDN) supports 8.5/15 = 57% of Geronimo committals, and a single commercial entity (JBoss Group) supports 30/85 = 35% of JBoss committals.

    There is nothing to suggest that Geronimo shouldn't be made to pay for certification too since it is largely developed by a commercial entity who intends to make money selling services based around the product. Whether that actually happens or not is another matter, of course.
  11. My dubious statistics[ Go to top ]

    Chris: There is nothing to suggest that Geronimo shouldn't be made to pay for certification too since it is largely developed by a commercial entity who intends to make money selling services based around the product.

    Like I said, set up a JBoss Foundation, give it the trade-mark, get broad buy-in and industry participation, and Sun will probably sponsor it. It's all about perception and politics, and JBoss Group is losing that battle right now because it looks like (perception) it can't decide between owning/controlling and opening/sharing.

    Other than some good-natured competition on a technological level, JBoss and Geronimo (and Jonas and Jetty and Tomcat and ..) shouldn't be butting heads .. remember, they both serve the same purpose, to save the world and bring app servers to the starving and disenfranchised. ;-)

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!
  12. What is JBoss anyway ?[ Go to top ]

    Maybe Marc you don't understand very well what "getting the certification" means.

    >
    > Suppose Objectweb would get the scholarship and Jonas would be certified.
    > ObjectWeb is a non-profit foundation. It's open to everyone. Jonas is LGPL. So what ?
    > Even JBoss could do business around Jonas. Even JBoss could join Objectweb.
    > What's inconsistent in that ? Certification applies to a piece of software, not
    > to a company. Scholarship applies to an organization, not to lines of code.
    >
    > The thing is, one never knows if "JBoss" refers to a company or a project. Do the ads up this page apply to JBoss-the-server or JBoss-the-company ?
    > Sounds like a clever little for-profit guy set up a company with a front open-source-but-not-open-minded project :)
    >
    > Suppose Geronimo in some remote future would get the certif too, Apache
    > getting the scholarship... The JBoss Group could just do the same and make some money from Geronimo.
    >
    > Red Hat did not join Objectweb to get Jonas : it's already open-source ! Just use your brain :-)
    > They put pretty clearly that they joined to donate some code and contribute to the development effort.
    > Objectweb was their only choice : the only application-server to be fully operational and to be developped
    > in a fully open way.
    >

    And how is JBoss not developed in a fully open way? You don't get to 85 developers with CVS access not being open. Please explain exactly as your comments I do not understand. My colleague got CVS access after a few complicated bug fixes because he asked.


    > Now let's imagine JBoss (the company, not the eponymous project) would get the scholarship.
    > Could anyone (Red Hat and al) get aboard JBoss (the project, this time) ? Well,
    > according to recent events, everybody doubts it. j

    Do you mean the CDN persons? If I started a successful open source project, I would be quite displeased if somebody forked it. If the LGPL and copyright violations are true, then this is even worse and should not be tolerated. Although they seem to like to think so, the CDN was not the JBoss project. People like Scott Stark and Adrion Brock are what really makes JBoss work. They have been so nice to me on the forums whenever I have had questions and fix my bugs for free.

    > Maybe this is the real difference between JBoss' status and Objectweb's.
    > Objectweb is an open non-profit whereas JBoss is not (neither the company nor the project).
    >

    JBoss is open-source and freely available. They have many CVS developers who are not employed by JBoss Group.

    > There's nothing to get mad about.
    >
    > As for me, I wouldn't offer to pay, since they are vibrant non-profit projects around, that may be certified soon, without anybody spending mucho dolares - and without you making your pile with my bucks and sweat.

    So, to be pure, you have to work late at night and have a "real" job to be open source? You have to be sponsored by Big Blue or Sun to be a full time open-source developer? It is good to see that it is possible to make a living at open source without having to result to sponsorship (Apache/Sun/Big Blue), or by dual licensing (MySQL), or by selling pre-packaged open source and polluting it with proprietary things(Redhat). I feel bad for the Jonas developers because they will be exploited by Redhat. Shame on ObjectWeb! Look out for your developers, don't sell them out.

    Hans
  13. What is JBoss anyway ?[ Go to top ]

    I think the difference in this case is that it's JBoss Group people who decide what goes into a release, and what the primary direction of each new release is. To put it succinctly, they control it - the for-profit JBoss Group employees control what gets released and called "JBoss Release X.Y.Z".

    A big part of the difference is strategic new feature implementation vs. bug fixes. On the bug fix side, Marc Fluery effectively has 60 or so people who will do the grunt work of fixes for free. On the strategic new feature stuff, I think you'll find that this is restricted very heavily in favor of JBoss Group employees. In this way JBoss Group gets a big array of free bug fixers, and an even bigger QA group that's also free, but maintains control of the strategic direction of the code base through his JBoss Group employees.

        -Mike
  14. What is JBoss anyway ?[ Go to top ]

    I think the difference in this case is that it's JBoss Group people who decide what goes into a release, and what the primary direction of each new release is. To put it succinctly, they control it - the for-profit JBoss Group employees control what gets released and called "JBoss Release X.Y.Z".

    >

    This is untrue (and would like to know what makes you think so). I do not work for JBoss Group, but do work on the JBoss-MX project. I have the freedom to work on the features that I wish and have the ability to release when I like. However, I usually communicate with the other developers on the project (some of which do work for JBoss Group) because it would be stupid not to. I am also co-lead on the JBoss-Remoting project, which is also, IMHO, very strategic to JBoss 4 (BTW, the other co-lead, does not work for JBoss Group either).
  15. What is JBoss anyway ?[ Go to top ]

    \Tom Elrod\
    This is untrue (and would like to know what makes you think so). I do not work for JBoss Group, but do work on the JBoss-MX project. I have the freedom to work on the features that I wish and have the ability to release when I like. However, I usually communicate with the other developers on the project (some of which do work for JBoss Group) because it would be stupid not to. I am also co-lead on the JBoss-Remoting project, which is also, IMHO, very strategic to JBoss 4 (BTW, the other co-lead, does not work for JBoss Group either).
    \Tom Elrod\

    It was the impression I had from looking at the committer list, seeing who was doing what sort of code changes, and getting a vague impression of the release process from poking around in CVS and elsewhere. Apparently my impressions were mistaken - thanks for setting me straight.

    Perhaps JBoss should publicize this sort of thing in some manner - it would certainly help their PR.

        -Mike
  16. TPC and x86[ Go to top ]

    Vic: Even TPC C show that "x86" Linux machines are much faster and much cheaper than Sun boxes.

    If only the world were as simple as you make it out to be ;-)

    For uniprocessor general-purpose servers, nothing can touch either the performance or the price of x86 commodity servers. For under $2000 you can get a very well equiped 1u x86 commodity server that is 4 or 5 times as fast as the Sun Sparc II uniprocessor servers. (Sun wasn't selling uniprocessor Sparc III servers last I checked.)

    For dual processor servers, it heavily depends on the workload that the server will do. That is because the x86 servers (other than the new AMD Opteron) have very limited bus architectures, so the two CPUs have to share the resources of a uniprocessor system. In other words, dual processor x86 servers are actually PCs with an extra processor, for all practical purposes. If the load is compute intensive, the scalability is close to 2x; if it is memory intensive, the scalability is significantly reduced, maybe to 1.6x or so. (Microsoft SQL Server scalability on TPC benchmarks on x86 is traditionally around 1.7x.)

    Sun has three lines of dual processor Sparc servers. The old Sparc II which is not worth buying now (except for low-power low-throughput hardened telecom stuff), the Ultra Sparc III which is a processor designed for 100+ CPU systems, or the Jalapeno ("IIIi" I think) which is more like a Xeon MP in its cache size and architecture and cost. Frankly, unless your app just magically happens to fit inside the 8MB cache of the USIII chip, it's not worth it for dual processor machines. The Jalapeno machines are about half as fast as the Xeons (maybe a little better -- the P4 Xeon has a very low IPC) and are typically more expensive.

    In other words, Sun's Sparc lineup will not compete much at all on the low end when it comes to price/performance.

    When it gets to four CPUs and beyond though, Sun gets competitive. That's because the scalability per Xeon CPU is (as above) often around 70%. These CPUs run about $4000 (actually, much more when you buy from HP or Dell etc.) So if a Xeon is over 1/2 as expensive as an USIII and less than twice as fast, it theoretically cannot be more than 4x as effective per dollar in a single CPU system. Since it only scales at 70%, so the fourth CPU is only worth 1/3 of a CPU (100%, 70%, 49%, 34%) and the eighth CPU is only worth 1/12 of a CPU (..., 24%, 17%, 12%, 8%). So compared to a 1 CPU Xeon box, the 8 CPU Xeon box is actually only doing about three times the work! The other surprise is that the price of the 8 CPU Xeon box is not eight times the cost of the 1 CPU Xeon box, but 25 or 50 times the price!

    Sun servers scale much better. If I remember correctly, for Java application server tests done by Ace's Hardware, they saw 96% scalability per CPU up to 64 CPUs. So let's re-run those numbers: 100%, 92%, 88%, 85%, 82%, 78%, 75%, 72%. So the eigth CPU in the Sun box provided better scalability than the second CPU in the x86 box! So eigth CPUs process almost seven times the work as one CPU! Plugging in some of the above assumptions, you'll find that the 8-way Sun box will out perform the 8-way HP (etc.) x86 box, and furthermore, if you have to scale up more, you can! And when it comes to pricing, Sun isn't much more expensive at this level than HP or Dell (they don't even have 8-ways) or IBM (probably cheaper than IBM).

    There are some curve balls though that threaten to upset this entire apple cart. First of all, the new AMD Opteron server chip will scale to 8 CPUs without any additional "glue", meaning that you could see 8-way servers for $20,000 (which is a fraction of what they cost today!) that scale almost as well per CPU as a Sun server and outperform the Xeon significantly even on a single CPU basis. Unfortunately, AMD has a big problem: It seems like they lose money every single quarter, which is a business disaster waiting to happen.

    Second, Intel built their own 64-bit server chip called Itanium, which finally in 2004 is performing as well as it was supposed to be in 1998 (when it didn't exist yet because the project was so far behind). If Intel can actually get people to buy it in quantity, the server costs around the Itanium will come down because it could become another "commodity market". Unfortunately, it doesn't appear to be happening, as most of the Itanium servers that have been "sold" were actually given away to big corporations, hoping that they would eventually buy more. (Not counting HP, which has bet the farm on Itanium, I think the total number of Itanium servers sold is only a couple hundred. Sun sells that many in a day!)

    If Sun were smart, they would have already bought AMD and they would be competing on performance AND on price/performance, and they would keep the USIII in the high end, and eventually merge for binary compatibility, much as what HP did with PA RISC into Itanium. Sun has great hardware engineers, they really do, and they finally (after 20 years) make hardware that works very well and is almost trusted for "enterprise computing" (something that Xeon is still struggling with.) Instead, they continue to lose the "low end", which continually grows higher and higher. "Low end" used to be uniprocessor ... now it is quite likely up to 8-way systems! That's getting into the "bread and butter" area for Sun. If they lose that, they're going to be another Cray selling only a handful of huge machines.

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!
  17. TPC and x86[ Go to top ]

    Vic: Even TPC C show that "x86" Linux machines are much faster and much cheaper than Sun boxes.

    >
    > If only the world were as simple as you make it out to be ;-)
    >
    > For uniprocessor general-purpose servers, nothing can touch either the performance or the price of x86 commodity servers. For under $2000 you can get a very well equiped 1u x86 commodity server that is 4 or 5 times as fast as the Sun Sparc II uniprocessor servers. (Sun wasn't selling uniprocessor Sparc III servers last I checked.)
    >
    > For dual processor servers, it heavily depends on the workload that the server will do. That is because the x86 servers (other than the new AMD Opteron) have very limited bus architectures, so the two CPUs have to share the resources of a uniprocessor system. In other words, dual processor x86 servers are actually PCs with an extra processor, for all practical purposes. If the load is compute intensive, the scalability is close to 2x; if it is memory intensive, the scalability is significantly reduced, maybe to 1.6x or so. (Microsoft SQL Server scalability on TPC benchmarks on x86 is traditionally around 1.7x.)
    >
    > Sun has three lines of dual processor Sparc servers. The old Sparc II which is not worth buying now (except for low-power low-throughput hardened telecom stuff), the Ultra Sparc III which is a processor designed for 100+ CPU systems, or the Jalapeno ("IIIi" I think) which is more like a Xeon MP in its cache size and architecture and cost. Frankly, unless your app just magically happens to fit inside the 8MB cache of the USIII chip, it's not worth it for dual processor machines. The Jalapeno machines are about half as fast as the Xeons (maybe a little better -- the P4 Xeon has a very low IPC) and are typically more expensive.
    >
    > In other words, Sun's Sparc lineup will not compete much at all on the low end when it comes to price/performance.
    >
    > When it gets to four CPUs and beyond though, Sun gets competitive. That's because the scalability per Xeon CPU is (as above) often around 70%. These CPUs run about $4000 (actually, much more when you buy from HP or Dell etc.) So if a Xeon is over 1/2 as expensive as an USIII and less than twice as fast, it theoretically cannot be more than 4x as effective per dollar in a single CPU system. Since it only scales at 70%, so the fourth CPU is only worth 1/3 of a CPU (100%, 70%, 49%, 34%) and the eighth CPU is only worth 1/12 of a CPU (..., 24%, 17%, 12%, 8%). So compared to a 1 CPU Xeon box, the 8 CPU Xeon box is actually only doing about three times the work! The other surprise is that the price of the 8 CPU Xeon box is not eight times the cost of the 1 CPU Xeon box, but 25 or 50 times the price!
    >
    > Sun servers scale much better. If I remember correctly, for Java application server tests done by Ace's Hardware, they saw 96% scalability per CPU up to 64 CPUs. So let's re-run those numbers: 100%, 92%, 88%, 85%, 82%, 78%, 75%, 72%. So the eigth CPU in the Sun box provided better scalability than the second CPU in the x86 box! So eigth CPUs process almost seven times the work as one CPU! Plugging in some of the above assumptions, you'll find that the 8-way Sun box will out perform the 8-way HP (etc.) x86 box, and furthermore, if you have to scale up more, you can! And when it comes to pricing, Sun isn't much more expensive at this level than HP or Dell (they don't even have 8-ways) or IBM (probably cheaper than IBM).
    >
    > There are some curve balls though that threaten to upset this entire apple cart. First of all, the new AMD Opteron server chip will scale to 8 CPUs without any additional "glue", meaning that you could see 8-way servers for $20,000 (which is a fraction of what they cost today!) that scale almost as well per CPU as a Sun server and outperform the Xeon significantly even on a single CPU basis. Unfortunately, AMD has a big problem: It seems like they lose money every single quarter, which is a business disaster waiting to happen.
    >
    > Second, Intel built their own 64-bit server chip called Itanium, which finally in 2004 is performing as well as it was supposed to be in 1998 (when it didn't exist yet because the project was so far behind). If Intel can actually get people to buy it in quantity, the server costs around the Itanium will come down because it could become another "commodity market". Unfortunately, it doesn't appear to be happening, as most of the Itanium servers that have been "sold" were actually given away to big corporations, hoping that they would eventually buy more. (Not counting HP, which has bet the farm on Itanium, I think the total number of Itanium servers sold is only a couple hundred. Sun sells that many in a day!)
    >
    > If Sun were smart, they would have already bought AMD and they would be competing on performance AND on price/performance, and they would keep the USIII in the high end, and eventually merge for binary compatibility, much as what HP did with PA RISC into Itanium. Sun has great hardware engineers, they really do, and they finally (after 20 years) make hardware that works very well and is almost trusted for "enterprise computing" (something that Xeon is still struggling with.) Instead, they continue to lose the "low end", which continually grows higher and higher. "Low end" used to be uniprocessor ... now it is quite likely up to 8-way systems! That's getting into the "bread and butter" area for Sun. If they lose that, they're going to be another Cray selling only a handful of huge machines.
    >
    > Peace,
    >
    > Cameron Purdy
    > Tangosol, Inc.
    > Coherence: Easily share live data across a cluster!

    That sounds long, Cameron, but what about tcp.org? The itanium system is roughly as fast as the HP or IBM stuff. And Sun Servers are not faster than the equivalent HP or IBM server.
  18. TPC and x86[ Go to top ]

    That sounds long, Cameron, but what about tcp.org? The itanium system is roughly as fast as the HP or IBM stuff. And Sun Servers are not faster than the equivalent HP or IBM server.

    I'm not an expert, I just stayed at a Holiday Inn Express .... ;-)

    The newest Itanium systems are actually taking top spots in certain benchmarks. The hardware architectures each have their ups and downs, in performance, scalability, cost and reliability. Itanium for example processes SSL about 10x faster than a Xeon, but processes 32-bit x86 code at about the speed of an old Pentium Pro 200Mhz (because it is largely emulated). Interestingly, when I priced Itanium systems six months ago, they were more expensive than Sun for equivalent performance, but that will probably flip-flop by this time next year. Last I checked, I could not actually buy the Itaniums that the new benchmarks were run on ... those are the (IIRC) 6MB cache 1.5Ghz chips, and they are both expensive and non-existent, but (again, IIRC) will be more easily available within a quarter or two.

    Regarding IBM, they definitely have a performance lead on Sun with the Power chips, in some areas over 50% faster (plus dual cores etc.) ... but they are typically two to four times as expensive, so you aren't going to save money buying IBM hardware (most companies aren't just buying the hardware, though, they have multi-tens-of-millions of dollars multi-year deals with IBM that makes the pricing a little closer to competitive.) IBM has a good name for reliability for enterprise computing, though ... most companies consider IBM hardware to be more reliable than Sun for example, and Sun is pretty good.

    I don't personally have a favorite. Intel/AMD (Dell, HP, whitebox, ...) in the "commodity market" are amazingly good. IBM and Sun (and I think HP again soon) in the high-end are amazingly good. What we get per dollar today is 100x what we got per dollar just several years ago, and it's because these companies compete hard for our business, and have really smart people coming up with (and often "borrowing" ;-) really nifty ideas. There are companies out there putting together 1000+ node computational grids that costs less than the older many-CPU servers that the grids are replacing, and will provide well over 100x the throughput. Similarly, you can buy a 100+ CPU server today and have it set up and running inside 90 days. These are things that we would have had a hard time believing if you suggested them ten years ago.

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!
  19. Open source all the way ;-)[ Go to top ]

    Vic> I see open source as "for-profit" and commercial as anti profit.

    I love that line! If you're in the open source boat, as your message implies, then you should tell whoever pays your salary that you no longer want it, that you're a "for-profit" guy who wants to get your corporate department on the revenue side of the chart of accounts. ;-)

    Seriously, open source has been a very good thing for the J2EE ecosystem. However, I don't believe it would thrive without commercial products. There are a lot of open source projects being developed purely by techs that live off of service $$$. However, there are also a large percentage of open source projects that are born, raised, and maintained by techs being paid by commercial companies. Some examples are XMLBeans, Eclipse, EJBGen, Grinder, etc.

    That doesn't mean that commercial products are better than open source products or that the more a product costs, the better it is. There are lots of examples to the contrary. It just means that the J2EE ecosystem cannot survive on service $$$ alone. If BEA, IBM, Sun, Oracle, etc. stop marketing commercial J2EE products, I think Microsoft would have most J2EE clients convinced to move to .Net within a year or two.

    Businesses don't choose commercial J2EE products over open source just because they have money to spend. They are paying for a perceived product stability, product support, version migration support, product training, and finger-pointing among other things. Unfortunately this last point is more important than maybe it should be. Of course companies could put together a "free" architecture composed of distinct open source products, and there are companies that do. However, the companies that choose a 100% open source architectura generally have higher-salaried technical staff capable of resolving the major integration and support issues rapidly in order to minimize the risks (both real and perceived) of such an architecture.
  20. We are JBoss customers, and aren't that really worried about certification. Sun's recent positions on Swing, Eclipse, and SCO make it known they are not a friend of the community if they can't control it directly. McNealy has pretty much gone off the deep end.

    Steve
  21. Objectweb's position[ Go to top ]

    http://www.objectweb.org/phorum/read.php?f=24&i=39&t=39
  22. Here is an excerpt of ObjectWeb's open response to Darryl K. Taft, the author of the article "Sun, JBoss Continue to Butt Heads Over Java" that appeared on eWeek :

    "[ObjectWeb has] been quoted in your article "Sun, JBoss Continue to Butt Heads Over Java" by a non-member of the Objectweb's community. We would like to correct two basic un-truths.

    [...]

    INRIA(*), as the host of the ObjectWeb consortium, is indeed applying for a Sun sponsorship via the JCP Scholarship. For the time being, INRIA has not yet received a licence from Sun.

    INRIA (being a public research institute) and ObjectWeb (an open international consortium) are not-for-profit organizations. ObjectWeb, in particular, does not sell its code base, which is freely available, without any discrimination, to anyone, for any purpose.

    [...]"

    (*) INRIA is better-known for hosting the W3C in Europe last years, in pretty much the same way it hosts ObjectWeb.
  23. Ramifications of JBoss redistribution...[ Go to top ]

    Who has something useful to say about the ramifications of building an application on JBoss to support internal business processes and/or redistribution for commercial purposes?

    Does Sun have a legal stranglehold due to licensing in either case?

    Would using JBoss be a legal risk to a company?
  24. Ramifications of JBoss redistribution...[ Go to top ]

    Who has something useful to say about the ramifications of building an application on JBoss to support internal business processes and/or redistribution for commercial purposes?

    >
    > Does Sun have a legal stranglehold due to licensing in either case?
    >
    > Would using JBoss be a legal risk to a company?

    Again, I heard that JBoss Group has agreed to the cash payout and contract and is just waiting on Sun for final stamp. I can't see how it is a legal risk if JBoss Group has agreed to the terms of licensing certification.

    Hans
  25. Maybe it's time for "The JBoss Foundation" to be formed, entrusted with the trademark, and with an independent board of directors including representatives from JBoss Group, Sun, etc. As a "true non-profit", it could apply for -- and would probably be granteed -- the necessary keys to certification.

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!
  26. Lets pay up and put this behind us![ Go to top ]

    Look, I like Jboss and am willing to pay towards the cost of certification. I am sure there are others like me as well. Jboss has a big community-- so lets all chip in and get this over with.
  27. Lets pay up and put this behind us![ Go to top ]

    Look, I like Jboss and am willing to pay towards the cost of certification. I am sure there are others like me as well. Jboss has a big community-- so lets all chip in and get this over with.

    Does anyone know how much money is being talked about here? I heard at one point that it was close to US$200M ... is that right? Can anyone close to JBG comment?

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!
  28. Lets pay up and put this behind us![ Go to top ]

    I heard at one point that it was close to US$200M ... is that right? Can anyone close to JBG comment?

    Only $200m ? Thats chump change for the Jboss community . After all it has been downloaded by a zillion developers !!! ;-)
  29. M&Ms[ Go to top ]

    Only $200m ? Thats chump change for the Jboss community . After all it has been downloaded by a zillion developers !!! ;-)

    Sorry, I should have been clearer: US$200M means "two hundred thousand dollars". In America, that's $200,000 and in Europe it's only $200.000.

    Peace,

    Cameron Purdy
    Tangosol, Inc.
    Coherence: Easily share live data across a cluster!
  30. M&Ms[ Go to top ]

    Good Lord, I am getting like the Yanks myself, I think.

    I read US$200M as "two hundred million" US :)
    Not as US$200K, which I would read as "two hundred thousand" US.

    Still...I would love to see them pay, and get the certification. It would at least silence the annoying habit of the BEA people pointing it out.

    Kinsley
  31. Lets pay up and put this behind us![ Go to top ]

    Look, I like Jboss and am willing to pay towards the cost of certification. I am sure there are others like me as well. Jboss has a big community-- so lets all chip in and get this over with.


    I heard that JBoss Group has been willing to pay for certification, has signed all the contracts, but Sun is again back to ignoring them.

    I think Sun is just sore that one, they can't make any money off of Sun ONE, and two, JBoss is the real RI.

    Hans
  32. JBOSSS: Really for profit![ Go to top ]

    What´s the big deal about Jboss been a licensed application server ?
    How about more profit, through more JBoss distribution, for JBoss Group.

    JBoss and JBoss Group relationship is not like Mozilla and Netscape one. JBoss is actually a JBoss Group product. If they charge for licences or support is up to them.

    As a profit company, The Jboss Group should pay for TCK license. Do they think Sun is a charity company ?
  33. Commercial does not mean bad.
    But if a company manipulates JCP, ASF, Linux (SCO), jBoss, they are anticompetitive (trying to limit open source); and since I have used open source for profit to pay my bills, I do not recommend Sun to my clients. As a libertarian, I do not want government assistance to make it competitive, since I can help it myself.

    As far as HW performance, I run App. servers. I have no formal test, but I know that running 8 1U x86 boxes is more scalable than a 16 proprietary CPU, when you run app. servers. (AND it is more affordable) It is silly to say, oh, you just configure 16 JVM processes for GC, etc. I hope no one is implying you run a single JVM accross 16 CPUs. It is much simple just to get 8 boxes, each 1U (ex: NewISys 2100)

    I think J2EE is a commodity now, because it is X-platform (GREAT!!!!). I think that is why Sun is trying to close it up, but it's to late. People that use jBoss I do not think care that Sun has not blessed jBoss. Who cares what Sun thinks. jBoss should just say that they are jBoss compliant, just like http://kaffe.org says in bold "Kaffe is not Java"! :-)

    .V
  34. For the record[ Go to top ]

    I just want to clarify our position since the article (as every article does) did not give all the context.

    The context is that JBoss Group has been pursuing certification from SUN since the late 2000 days. We made official request through our lawyer to pass certification. SUN never responded until this spring of 2003, right before Java ONE when they approached us with a TCK offer.

    I can't disclose the sum, let's just say high 6 figures (in US dollars) and made it clear that JBoss Group, as a commercial entity, could only get TCK for pay.

    WE AGREED TO ALL THE TERMS back in july, including the money. The money is a hefty sum for us. We put out a PR announcement that we had agreed to everything, all the terms, and were waiting on SUN to send the contract.

    We are still waiting. It is a long string of "next week, next week" for the past 8 weeks from SUN sales guys.

    Meanwhile we hear that the scolarship request from ObjectWeb has been approved.

    WE ARE NOT OFFENDED BY THE FACT THAT WE SHOULD PAY, WE OFFENDED BY THE LACK OF CONSISTENCY FROM SUN. We have never pretended to be a non-profit, we are not a non-profit and we believe in professional open source.

    What we feel is unfair, is the way the program is set up. Essentially you can set up a front non-profit and get the certification for free, even though you are France Telecom, Group Bull and RED HAT. While the 30 developers at JBoss have to pony up mucho dolares. Read the article where I argue "proof by the absurd".

    Finally if you parse "SCHWARTZ" language you see that he says "people who don't charge for the product get certification for free" (first paragraph). Well we would qualify in that condition... so what is going on?

    I think there is a lot of confusion, legitimate confusion, over at SUN on how to deal with an OPEN SOURCE COMPANY. We have excellent relations with IBM, but SUN can't figure out the way to do it and their first attempt, that 'scolarship program' arrives at non-sensical, non fair, non consistent results.

    My argument is "back to the drawing board". If SUN wants to support Open Source they should start with licenses, which is the real definition of open source. We hope to solve this issue with SUN.

    Thanks for those that offer to pay :) that's nice, we will pay, but we want consistency and fairness.

    marcf
  35. For the record[ Go to top ]

    more clearly

    EITHER WE ALL PAY, OR NONE OF THE OPEN SOURCE IMPLEMENTERS PAY.

    That's consistency.

    marcf