Open Source JBOSS J2EE Server v2.4 goes final

Discussions

News: Open Source JBOSS J2EE Server v2.4 goes final

  1. The Open Source J2EE server JBoss has posted its finalized 2.4 release for download.

    While peaking around the Jboss forums, I got a tip that JBoss 2.4.0 was now final. Good news indeed, but no one is talking about what is in it. What I would like to know is the extent to which they have supported EJBII in this release. One could experiment with things and find out what works and what doesn't. Anyone know anything about the CMP in 2.4?

    Download JBoss 2.4.

    Threaded Messages (60)

  2. If you click here, you will see the changelog notes for JBoss 2.4.

    Some of the highlights from that changelog are:

    JBossAdmin: Monitoring client
    Monitor client added from the admin module to JBoss distro.
    Provides basic monitoring of invocation times and cache usage.

    Made log4j the default logging framework
    The org.apache.log4j framework is now the default
    logging framework.

    CMP finder optimization
    Finders returning collections of entities are now
    optimizable in JAWS. This feature is activated by an
    optional <read-ahead> tag in the finder's configuration
    in jaws.xml.

    Transaction Isolation is configurable
    The JDBC connection transaction isolation level is now
    configurable with org.jboss.jdbc.XADataSourceLoader in
    jboss.jcml
     

    Hope this helps.
  3. and also look at this..
  4. It still does not say anything about EJB2.0 support. Can somebody tell me to wha level they have supported EJB2.0 PFD. Anyway, if somebody wants to play with EJB2.0, there is excellent product out there called "Pramati". Although this is a commercial product, but on their site one can download the alpha of Pramati 3.0 it has complete support for EJB2.0 final draft. And the evaluation license is valid for 60 days. I looked at it and seems to be a very promising product. It has a very well conceived UI etc. One can go to www.pramati.com.

    Vimal
  5. Message Driven Beans and business methods on the home interface are supported in JBoss 2.4
    The CMP 2.0 work is now in the main branch of JBoss and should be released in the next stable release.
  6. There is limited if any support for EJB2.0 in JBoss 2.4. That stuff is going to be in the 3.0 release which is supposedly going alpha before the end of the month.
  7. That is a great news. I will be waiting rather eagerly for this.

    Vimal
  8. JBoss without clustering[ Go to top ]

    Sorry for my previous rude comment. For the project that I am working on (only 5 developers), any application server without clustering support is not in our radar screen. I am an independent consultant working for a major telephone company. Our project is mission-critical and should be well-scaled beyong more than 500 concurrent users.

    Any J2EE application server should support distributed computing. If it does not, why should we use it at all? I have finished several JSP/Servlet/Bean projects using only servlet engines. These servlet engines are much cheaper to deploy and maintain. Only for these projects who have huge growth potential but currently not in much use may consider JBoss.

    I agree that JBoss is well-designed and read a note on Jboss.org that its lead developer is going working full-time to make this product clustering. That's why I am asking this question.
  9. JBoss without clustering[ Go to top ]

    James Liu,
    If you want clustering, but without full scale EJB or J2EE, take a look at Rexip AppServer (www.rexip.com)

    - Pete
  10. JBoss without clustering[ Go to top ]

    Jeff,

    As for scalability, JBoss was tested at Suns benchmark labs, and was found to scale nicely on their hardware (which was admittedly heavyweight, but not the best possible) to 5000 concurrent calls, each taking 9ms each. This is of course only for vertical scalability (i.e. how far you can get with one box) but should be ok for most users.

    /Rickard

    ps. For the whole story see:
    http://www.mail-archive.com/jboss-user@list.working-dogs.com/msg04427.html
  11. JBoss without clustering[ Go to top ]

    From the author of the "JBoss is a toy" wisdom comes:

    > For the project
    > that I am working on (only 5 developers), any application
    > server without clustering support is not in our radar
    > screen. I am an independent consultant working for
    > a major telephone company. Our project is
    > mission-critical and should be well-scaled beyong
    > more than 500 concurrent users.

    > Any J2EE application server should support distributed
    > computing. If it does not, why should we use it at all?

    Boggle, consultants w/o clue. Clustering is not distributed computing. Distributed computing is independant processes cooperating across multiple machines. See JNDI/RMI/IIOP.

    This clustering you seem so keen on is load-sharing/redundancy for the same processes using multiple machines. $20 says you've got your whole system running off one router, one db, or all on one rack under a sprinkler anyway.

    I have begun to notice that the less intelligent the people involved the more likely alot of money is being spent on big ticket commercial products. Compensation for having no talent in-house. If you have staff with a clue then JBoss (well open-source in general) is the ticket. On the other hand if your consultants are under the impression that clustering is distributed computing then best play it safe and call your bank manager for a loan to kit out with the $30k per CPU shrinkwraps.

    Dave
  12. JBoss without clustering[ Go to top ]

    Doesn't JBoss run on top of Tomcat, which is well known to be very slow, clustering or no clustering?

    How does JBoss compared with Orion?
  13. JBoss without clustering[ Go to top ]

    <i> Doesn't JBoss run on top of Tomcat, which is well known to be very slow, clustering or no clustering? </i>

    JBoss can <i>run with</i> Tomcat as the servlet engine. It can also run with Jetty as the servlet engine, which has the reputation of being faster.

    Tomcat isn't a speed demon, but I'm unaware of it being a complete pig... UNLESS it's improperly configured, that is.

    As far as serving up enterprise features (i.e. EJB's) goes, JBoss handles that part of the equation itself and does not reply on Tomcat (or any servlet engine) for those services.
  14. JBoss without clustering[ Go to top ]

    Dave Elliot,
    According to your definition of distributed computing, clustering is indeed distributed computing.

    They are independent processes serving different requests.
  15. JBoss without clustering[ Go to top ]

    Ah that often misused over-hyped word "clustering" strikes again.

    In this instance when we are talking about clustering, we are talking about replication of session data between different application server instances running on different Java VMs to achieve scalability and fault tolerance.

    Whether or not this is "distributed programming" or not, it is good stuff in theory.

    Different companies have different priorities that drive there decision to buy different application server products. This is a good thing. The only bad scenario is when a company buys into an application server based on market hype without the proper research.

    BTW, please don't waist everyone's reading time calling each other names. No one is an idiot here.
  16. JBoss without clustering[ Go to top ]

    I would argue that clustering _is implemented_ using distributed computing. If you're that keen to delve into a debate on the semantics of the matter please feel free to drop me an email ;-)

    In any case my objection was to the original poster who asserted that:

    "Without clustering support, JBoss is still a toy.", concluding that, "Any J2EE application server should support distributed computing. If it does not, why should we use it at all?"

    To claim that any J2EE server (or any Java based server for that matter) does not support "distributed computing" is, well, bollocks. Apologies to those who found my objection to these statements a touch strident. However I reserve the right to object strongly to the spreading of anti-open-source FUD whenever it should raise its ugly head. Next?





  17. JBoss without clustering[ Go to top ]

    Just for your information, I don't make decisions on which EJB server I should use. The telephone company that I am working for has made the decision. This company has a company-wide license with BEA.

    I understood your decision to use hardware to do the load-balancing. If you have to use session, you have to use the sticky-mode to do the hardware load-balancing. There are advantages and distadvantages on this decision. We had a debate on this issue. Eventually we chose BEA because BEA is "free" for the team.
  18. JBoss without clustering[ Go to top ]

    Indeed
  19. JBoss can work with a third-party product which basically takes EJB 2.0 and converts it into 1.1, allowing you to start writing 2.0 now and using them in JBoss as is.

    Search the mailing lists at jboss.org for more info.

    Peter.
  20. Any clustering support?[ Go to top ]

    Without clustering support, JBoss is still a toy.
  21. Any clustering support?[ Go to top ]

    That's a very silly comment.
    Do you really think that all J2ee applications need to be clustered?
  22. Any clustering support?[ Go to top ]

    Without clustering support, JBoss is still a toy.


    Though I'm using WebLogic I did not need clustering so far. So I guess my team of 20 developers is just toying around, right?

    Maybe you're in one of those big 5 consultancies that do not start a project below some millions of dollars, but I guess there are some hundreds of thousands of J2EE developers out there who are not in such a situation. And then a free quality server might be a great choice.

    Daniel
  23. Any clustering support?[ Go to top ]

    I completely disagree - but regardless - I believe they'll have clustering in the next major release.
  24. Any clustering support?[ Go to top ]

    I have used Weblogic before and I did not have any need for clustering. In my current project I have switched to JBOSS and I am very happy with it.

    My client did not have to pay for those overpriced App Servers. I don't have to make those frustrating phone calls to Tech Support, wait for days for a reply with answers to my questions and then finally figure out the solution myself. With JBOSS, I simply lookup the source code. Last but not the least, JBOSS has a very nice and elegant architecture.

    Pranab
  25. Any clustering support?[ Go to top ]

    Can you provide some details on what was involved in migrating from WebLogic to JBOss? Also what your configuration?
  26. Any clustering support?[ Go to top ]

    Robert,

    EJB deployment descriptor extensions are very similar to Weblogic. To deploy simply copy the jar files to JBOSS deploy directory. Hot deploy works too. It detects newer jar files in it's deploy directory and redeploys them.

    Other configuration files are XML based and pretty easy to deal with. One nice feature of JBOSS is that uses JMX heavily which allows it to have a pluggable architecture.
     
    Most of the internal components are implemented as JMX MBeans. They can be replaced if necessary. Additinal features e.g., initialization at startup etc are easy to add. You just create a MBean and define it's configuration in jboss.conf file.

    Hope it helps.

    Pranab
  27. Any clustering support?[ Go to top ]

    Pranab, that's a load of crap re: "I can look up the code and modify it". There's a simple reason why many people fork out the $$$ to pay for tech support. Enterprise applications depend on 24x7 support and not some hotshot like you who modifies code on the side.

    Get a life!
  28. Any clustering support?[ Go to top ]

    Huh?
    How much useful support have you ever had from them on the phone? Personally, user forums for any (reasonably used) product are infinitly better than calling a help desk who has know Idea.
    I love sitting on hold, and then waiting 5 days for the following work around to a fault:

    "You are correct, it is a fault. The solution is a documentation change"
    "Okay, what is it"
    "No longer supported in this configuration"
    "Great"
  29. Any clustering support?[ Go to top ]

    Tim,

    you're right, it's always nice to get answers from the forums, especially if the developers hang out there.

    But I don't share your point paid support. I've WebLogic paid support for each installation and I must say that they are really helpful. Until now they where able to deliver a solution in most cases and actually saved me a lot of work. I made the experience that on all kind of mailing lists people are really helpful when it's about explaining higher level problems, e.g. "how to configure this and that in my ejb", but certain problems that are somewhere deep in the server are not really likely to get answers on mailing list. As these non-standard problems are the real show-stoppers - most other questions on mailing lists *are* answered somewhere already and can be solved by carefully reading either in spec. or the server documentation - I see a real need for vendor support.

    Cheers,

    Daniel
  30. Any clustering support?[ Go to top ]

    Hi Daniel,

    You wrote:
    > but certain problems that are somewhere deep
    > in the server are not really likely to get
    > answers on mailing list. As these non-standard
    > problems are the real show-stoppers - most
    > other questions on mailing lists *are*
    > answered somewhere already and can be solved
    > by carefully reading either in spec. or the
    > server documentation - I see a real need for
    > vendor support.

    For such cases you can buy support for the JBoss server. See http://www.jbossgroup.com/JBG/support.jsp

    -- Juha
  31. Any clustering support?[ Go to top ]

    George,

    You did not read my posting carefully. I said I could I could read the code to get my questions answered. As far as modifying, I prefer to follow whatever open source procedures are in place and not make any ad hoc modifications.

    As far as getting a life - It's much better since I switched to JBOSS

    Pranab
  32. Any clustering support?[ Go to top ]

    Posted by George Art 2001-08-23 19:12:15.867.

    >Pranab, that's a load of crap re: "I can look up the code >and modify it". There's a simple
    >reason why many people fork out the $$$ to pay for tech >support. Enterprise applications
    >depend on 24x7 support and not some hotshot like you who >modifies code on the side.

    >Get a life!

    George -
    You seem to have missed the point of original posting. In open source software you can go and look at what the code is doing and either work around it accordingly, submit it to the development group as a bug, fix it yourself, etc. In my experience, $$$ tech support has almost never been worth the $$$ (with some exceptions - none of which have been J2EE app server products).
    Cheers

    Ray
  33. Any clustering support?[ Go to top ]

    I would like to know how to configure oracle for jboss?
  34. Any clustering support?[ Go to top ]

    I do not think that clustering support is over-rated technically. I think that anyone can fairly easily build a (*portable*) clustering solution into their web app using any app server (without requiring any costly and proprietary clustering support).

    There are at least two options.

    1) If you are able to design your application as being stateless (which generally means not using the HTTP session), then
    - Use hardware/software load balancing. That's it, there is your clustering support:-)

    2) If you need to maintain state information in the session, then
    - Use hardware/software load balancing with sticky sessions configured.
    - Use a cookie to persist session state changes to the database. This can actually be highly optimized so that session data that is only serving as cache data need not be persisted. In my experience this leads to very small amounts of infrequent database writing. In most cases only the "user id" of the person associated with the session needs to really be persisted (and this only occurs once per session).
    - Now the box goes down, the load balancing software redirects the request a new box (failover), a new session is created, we use the cookie to restore the session data from the database, and the lost cached data can be lazily reconstituted using the restored session data (eg. cached shopping cart data can be restored using the 'user id').

    I find more "sophisticated" clustering mechanisms like in-memory session replication to be less reliable, more complicated to configure, less portable, more costly, and slower. In-memory typically requires constant multicast network traffic (significant overhead as you scale), it's generally not transactional, and the synchronization protocols are not always well defined.

    Marko.
    .
  35. Any clustering support?[ Go to top ]

    Marko

    You're absolutely right about clustering in most circumstances. Besides the multicase heartbeat, the primary server needs to update all the others when the session state changes. So there is a performance hit for each session state update. Also all of the servers must be on the same subnet since multicast can't travel beyond that. BEA WebLogic does offer some efficencies by synchronizing only content that has changed but there are some significant limitations that need to be dealt with in your application.

    The clustering version of WebLogic is considerably more expensive than the non-clustering server by about 70% according to the last quote. That's per CPU.

    Anyway this comes down to the fact that in most situations you can do it easier and better with hardware load balancers than with software.

    Bob
  36. Any clustering support?[ Go to top ]

    Has anyone tried clustering of Rexip AppServer (www.rexip.com)? Does it work well?
  37. Any clustering support?[ Go to top ]

    Bob,
    I believe hardware load balancer complements software load balancer rather than replaces it. You cannot store the Java session data in the hardware load balancer, for example.

    Harware load balancer directs traffic to Web server farm.
    Web server farms direct traffic to Servlet server form...
    and so on.

    At each level, you can have clustering.

    I am curious. Has anyone done any benchmark comparing the "sophisticated" heartbeat in-memory replication compared with the more traditional approaches?

    Multicasting on an Ethernet will not consume more network resources (Even point-to-point traffic is broadcast on Ethernet anyway), but may consume more CPU resources. However, the keep-alive messages multicasted are really very small. The session data, according to my understanding, are not multicasted to everyone in the cluster. I will be surprised if in-memory replication does not stand up to the more traditional approaches of storing session data in database. But then again, only actual benchmark tests will tell.


  38. Any clustering support?[ Go to top ]

    Dan

    The software I was referring to was the clustering solution provided by app servers such as WebLogic. Of course in addition to using load balancers the session state must be stored in some manner as in the HTTP session, databases or in the case of clustering app servers using in-memory replication.

    The question is whether or not a clustering support was essential for a app server to be considered an enterprise solution or just "a toy" as someone mentioned. My point is that there is no free lunch. Even with clustering there are important considerations that need to made in your application about the particular clustering environment.

    Another point made was that clustering overhead is very small. That's only true if the contents of the session is very small. Otherwise changes in the state may force all of the session contents to replicated to each server which can be drag down the network.

    Bob
  39. Any clustering support?[ Go to top ]

    I am not sure why a few of you think in memory replication of sessions is so hard on the network? Orion app server does it right. They replicate only objects as they are set in the session. Sure, you have to add the setAttribute() for any object you "modify" of the session, but this ensures the optimal in-memory replication. I create the Theseus MVC framework and added support to get any bean it uses and put it in the session for you at the end of the request, thus avoiding having to program it yourself. Other objects would still require developer intervention.

    But all this talk about using multi-cast is network hogging and it has to replicate to all servers is not necessarily true. There are many ways to do this. The way I will incorporate into Jetty is much like Orion's method. Replicate only the object being stored. But in my docs explaining how to set it up, I will insruct that you use only two servers per cluster, using multiple clusters to scale. In this way, you only replicate session objects to one other server. Should one fail, the other becomes the "only" server until the dead one comes back up or is replaced. Using a free tool like Big Brother to monitor hardware can alert anyone to the problem right away and in a matter of hours a server can be replaced. You have to plan for this stuff though. If you have a high-traffic site, you of course set up a few extra servers (or clusters if you require session state fail-over) "just in case". Mission critical companies should never have a problem adding a few extra servers to make sure they are always up. And the stuff about network traffic. Come on..first of all, you use two network cards per server. One for the internet or main lan connection, and another for the back-end connection for fail-over. In other words, if you use two servers per cluster, you hook one of the two network cards up to a hub and in that hub is only two cables, one from each computer. If you want, you can even use a single cable between the two network cards (forget what its called where you use one cable to network only two computers). But more so, gigabit networks are not that expensive any more. Mission critical should always use gigabit. At $250 or so per network card, $500 for the replication network that gives you 100MB per second is pretty damn fast! If you had 500 concurrent users per server in a 2-server cluster, and each one is replicating one object per user per second, your talking about a mere 1000 or so objects per second transferring back and forth. I think the network can easily handle that, even at 100Mbps unless your transferring very large objects.

    The use of persistent store for sessions adds another point of failure, another server for storage, and of course the hd(s) its stored on as well. Sure, it allows X number of servers in a single cluster to all share the one session, which is nice, but you could potentially do something like this with in-memory where you have one server use nothing but memory (and perhaps OS page swapping of memory to the local HD which is probably faster than a connection to the database and lookups when getting the session state back out from the database) and have the servers in the same cluster hit that machine for their session state. This would require some special programming for the in-memory central server, but it is certainly possible.

    But I agree also. First, I think its best to use HttpSession for state, and use stateless EJB. Not sure why you need Stateful ejb at this point, but I only do web development. Maybe clients can use it? No idea. More servlet containers are available for session fail-over clustering. EJB containers have a lot more to handle in terms of services it supports or is supposed to support. I think have x number of web-tier clusters (failed over if need be in 2 servers per cluster) hit a single hardware load balancer that hits x number of EJB servers in the middle tier that don't need to be clustered provides a rapid scalable architecture for both tiers. Anyone care to comment if this is good or not?
  40. Does anyone know of a specific large scale enterprise installation of JBoss as the primary app server? Can you provide details on the configuration, scalability, uptime, support and ease of adminstration?

    BTW I downloaded the 2.4 release. It installed and started up without incident. I'm going to do some simple head to head comparisons with BEA WebLogic 6.1.

  41. Hi Robert,
        We are currently considering seriously switching from Silverstream to JBoss. We setup a some comparisons and unfortunately the transactions per sec don't even come close to silverstreams transactions per sec , without getting into too much technical stuff we are using the same code base and I'm quite sure its something to do with configuration. We have posted numerous times to Jboss forum with loads of replies of which we have already tried most and rest didn't too much difference. Anyway I was hoping you might drop me a line on major config issues you come across and how its performance compares with Weblogic.

    Thanks very much
    Paul B

  42. Can you describe in some detail what your transactions consist of? Have you tried other app servers?
  43. Rob:

    You will find JBoss running faster than WebLogic 6.0

    Tieu Chu
  44. You will find JBoss running faster than WebLogic 6.0


    Can you post the real numbers? where does stats come from?

    Thanks
  45. Well I'm only going to do the comparison against WebLogic 6.1 since BEA claims it's the fastest app server on the market. Supposedly much more optimized than 6.1

    I will report the findings.
  46. Well I'm only going to do the comparison against WebLogic > 6.1 since BEA claims it's the fastest app server on the

    > market. Supposedly much more optimized than 6.1

    > I will report the findings.

    good luck dude! If you read the license terms carefully, most companies expressely forbid you from releasing any benchmark numbers without their explicit permission.
  47. good luck dude! If you read the license terms carefully, >most companies expressely forbid you from releasing any >benchmark numbers without their explicit permission.


    ... which may not be lawful in many countries - isn't that covered by "freedom of speech"?. Beside that it's really ridiculous. Just imagine Porsche sueing people who claim that a Ferrari goes 5 mph faster than a Porsche ...
  48. <quote>
    ... which may not be lawful in many countries - isn't that covered by "freedom of speech"?.
    </quote>

    What do laws passed by the Congress have to do with this thread? Because this is probably the "freedom of speech" you are talking about.

    There is no absolute freedom of speech. Fortunately, or our world would be total chaos otherwise.

    For example, do you think you are allowed to post internal information about the company you work for because of "freedom of speech"?

    There are contracts, license agreements and other legal binding documents that restrict your freedom. You are free not to sign them, but if you do, you purposefully give away some of your freedom of speech, typically in exchange of tangible benefits (such as a salary).

    Freedom of speech is a concept that has been vastly exaggerated :-)

    --
    Cedric

  49. About this earlier comment:

    ...
    good luck dude! If you read the license terms carefully, most companies expressely forbid you from releasing any benchmark numbers without their explicit permission.
    ...

    Legally, anyone can publish the results of a comparison of two or more products as long as an explanation of how the comparison was done is provided also. The results of any test are always judged by the validity of the test conditions.

  50. <quote>
    Legally, anyone can publish the results of a comparison of two or more products as long as an explanation of how the comparison was done is provided also. The results of any test are always judged by the validity of the test conditions.
    </quote>

    No.

    Legally, you're only allowed to do whatever is on the license agreement you signed.

    --
    Cedric
  51. i just download your JBoss J2EE Server v2.4, but in docs, it said the version is jBoss 2.o server, I am not sure if the version of this server is correct or not.
  52. Hello, many of you have posted questions regarding JBoss's performance and scalability, and usage. I am the J2EE "architect" (whatever that really means) for the Northwest Alliance for Computational Science and Engineering. We are a research group that works with everything from supercomputing to User Interface development and design. I primarily work in the middleware side of things doing research in distributed computing, distributed data management and mining, data interchange, that sort of thing.

    Here we have run analyses on quite a number of J2EE app servers in terms of many issues. Since we do work quite a bit in high-performance computing, performance is an issue (that sounds almost like a really stupid thing to say). One of the most important things is PRICE AS IT RELATES TO PERFORMANCE. This doesn't just mean the server software price (because then JBoss would be unbeatable). But a combined set of factors is important.

    For example: If I can put JBoss on a server with the following Sun Enterprise 450 Server:
    4 480MHz UltraSPARC-II Processors
    4MB External Cache
    4GB Memory
    4 36-GB 10000 RPM UltraSCSI Disk Drives

    Then I've got some heavy-weight power driving it. The total cost of ownership for this with JBoss is (I'm estimating) about $75,000 (actually less, but I'm overestimating by about $17,000 - $20,000 for fear of hidden and unknown costs).

    The same thing from Dynamo, Weblogic, Bluestone (HP), IBM, Silverstream would be hugely expensive. In some cases prbably more than ten times as expensive. Even if these systems beat JBoss on performance (which was not always the case) the PER UNIT COST OF PERFORMANCE is the key. For ten times the money, you don't get anywhere near the perfomance. WAIT, before you get mad at me read the next paragraph.

    OK, so lets say fully HALF of the price of the commmercial apps is for more features, support, and all of the other stuff you get from the biggies. That still means that the price-to-performance ratio is still way in favor of the little guys. By little guys, I don't even mean just JBoss anymore. I mean systems like Orion and JRun. These are great systems that cost nothing when compared. For example the license cost for JRun Enterprise on the server above would be expensive compared to JBoss (about $7,000 more per CPU), but a drop in the bucket compared to the "biggies."

    One of the consistent reasons that I've seen companies use the biggies for is the extended functionality. And this does make VERY GOOD sense. Some people just want the functionality.

    Also, a lot of companies can't afford the cost of hiring a lot of extremely high level, highly paid developer architects. Buying a system like Weblogic, IBM, ATG makes sense for these people. A lot of the granular work can be cut out with the proprietary stuff, and a lot of companies don't worry/care how the proprietary stuff impacts portability etc. People do tend to need to be more knowledgeable when using something like JBoss (or open-source in genereal). There are no training courses, no support call centers etc. They do tend to incur the heavy one-time expense of the licenses, and far greater systems management expenses.

    The proprietary stuff for me is in fact just as prohibitive as the price. I'd rather do more development and stay a little more "standardized" if such a thing really exists.
    I however do have that luxury. I also don't have over a million to spend on servers. My boss would laugh so hard she'd faint if I told her we wanted to spend that kind of loot. She'd probably ask me if I was smoking crack or something.

    Lastly as a matter of point, the Northwest Alliance has tentatively selected JBoss as its J2EE server (combined with Tomcat for Servlet/JSP). Who knows, someday maybe we'll switch, but right now it's JBoss. We are also eagerly awaiting other open-source projects like Richard Monson-Haefel's project Exolab and Enhydra Enterprise. And maybe other factors will, over time, become important to us, like functionality and support.

    There are many reasons to use Weblogic (or biggies). And there are many reasons to use JBoss. There seems to be a lot of fighting over which is better. Neither is better when you take in all of the reasons that individual companies use for purchasing the one they do. In these conversations we tend to fixate on one issue, when there is never really just one issue.

    Instead of firing shots, why don't we all agree that most of these products are good for the right criteria. Let's all support J2EE in general, and the server vendors in specific.

    As a last matter, I have worked with both commercial and open-source systems. I've liked and hated them all. Each one has its frustrations and its joys. I am not an open-source junky. I'm fact I've never contributed to an open-source project. I'm just a leech.....I am, hoever, a firm believer in "Pick the right tool for the job at hand!" And do this without prior bias. When I've done this, projects have been more successful. When I've lost sight of this, projects were'nt necessarily doomed, but they WERE harder.


    Sorry for the long post. And please don't flame me.

    -Newt







  53. Newt - Right on. Anything can be right in the right place.

    Jim
  54. Hi Jason,

    > People do tend to need to be more
    > knowledgeable when using something
    > like JBoss (or open-source in genereal).
    > There are no training courses..

    Actually, we do have training for JBoss available. The next sessions will be held in September (London) and October (Las Vegas).

    For more information see the jboss.org website and click on the training link on the left.

    -- Juha
  55. Can JBoss be made to run on faster servlet containers than Tomcat, like JRun, Rexip, or Resin. Besides faster performance, Rexip also claims to support clustering (servlet clustering, not EJB, which JBoss lacks currently).

    Would combining JBoss EJB with another servlet container make sense - or it is not possible?

  56. JBoss is not tied to any particular Servlet engine. There are prepackaged downloads for both Tomcat and Jetty, but others are certainly possible. There are certain benefits in running the Servlet engine from the same JVM as JBoss, but there is no restriction that forces you to do so.

    In fact, the beauty of the distributed model is that you can run the JSP/Servlet engine seperately from the Application Server itself...this lets you use any Servlet engine you care to.
  57. I am working with a small non profit company with a 1.1 Ghz 512MB RAM Linux server. Currently we are developing with Orion, but are looking into JBoss considering that $1500 is still a lot of money for us. After reading the above post about the $75000 Sun server and almost fainting, I am curious if the app server we currently have can do a good job on <100 concurrent requests?

    Thanks,
    Curtis
  58. I believe orion is free for non commercial use. So you can still use orion for your non profit company.

    Can't comment on your machine except that you should probably increase the RAM a bit more, say 1G. :) Just MHO.
  59. Curtis,
    Do you use EJB? If not, consider tomcat, rexip, jrun, resin, or other servlet container?
  60. Yes I am using EJB in the form of Entity CMP beans and session beans.

    Curtis
  61. Hi,

    I would like to know whether JBoss 3.0 supports, Session Distribution and Clustering with Load Balancing between different physical servers. I have tried to find out it from JBoss forum at http://www.jboss.org but its down.

    Any help would be appriciated.

    - Lomesh.