Discussions

News: J2EE market shares revealed for 2001

  1. J2EE market shares revealed for 2001 (57 messages)

    A report from Giga Information Group unveils the J2EE market shares for 2001. According to that report: "The application server market grew 39 percent to $2.19 billion during 2001. IBM and BEA Systems share market leadership, each with a 34 percent market share - leading closest competitors with a 68 percent combined share. Although Sun/iPlanet retained the third spot, Oracle's 9iAS grew faster to reach a 6 percent share and appears to have a higher run-rate during the second half of 2001."

    Full articles are available at:
    - http://www.gigaweb.com/browsemktg/0,2973,strComp%25253Dgigaflash,00.html#GF4
    - http://biz.yahoo.com/djus/020321/200203211614000597_1.html

    This report shows several interesting things. First, J2EE is in very good health, which is *the* good news. Then BEA, despite aggressive attacks from both IBM and Oracle, is still the J2EE leader (contrary to Oracle's very serious -giggle- survey at http://www.oracle.com/features/index.html?t1as_mktshare.html). We've seen many passionate debates here between religious WebLogic/Websphere zealots after recent IBM's ECPerf results, then between .NET/J2EE supporters. But I guess the real challenge will be to keep up the market shares for 2002 with .NET entering the playground.

    Any thoughts ?

    Threaded Messages (57)

  2. How I rank the J2EE platforms[ Go to top ]

    Based only upon my personal experience, here is how I rank the J2EE app servers:

    1. Resin
    2. Orion
    3. Oracle (aka Orion)
    4. BEA
    5. IBM
    6. JBoss
    7. HP-AS (aka Bluestone)

    How does your list go?

    *Flame Shields are at maximum* :)
  3. How I rank the J2EE platforms[ Go to top ]

    Based on my personal experience, I rank the app servers as follows:

    Orion
    Oracle
    EA Server (4.1)
    BEA
    JBoss
    Silverstream
    HP-AS
  4. How I rank the J2EE platforms[ Go to top ]

    1.BEA
    2.IBM
    3.HP-AS (aka Bluestone)
    4.Oracle (aka Orion)
  5. J2EE market shares revealed for 2001[ Go to top ]

    BEA doesn't look like a J2EE leader according to posted marketshare figures.
  6. J2EE market shares revealed for 2001[ Go to top ]

    The market share figures are very skewed to favor a company like IBM.

    It's based on percent of revenue- so while BEA may have 70% of the total servers in the market, IBM has an equal amount of revenue and thus they are equals in this survey.

    That being said the space is evolving into value added: Security, single-sign on, personalization, content management, flexible (proprietary) rapid application dev, Web Services, Portlets, vertical components, EAI etc.

    The only app server players that will be in this market in 3 years are:
    Oracle
    Microsoft
    HP
    Sybase
    BEA
    IBM
    Sun
    Borland (Maybe)

  7. J2EE market shares revealed for 2001[ Go to top ]

    Based on my way to evaluate application servers' success (go to www.jobserve.com, type your favourite application server name and hit search, you'll get a number of vacancies):

    - .NET: 404
    - Weblogic: 219
    - Websphere: 177
    - iPlanet: 42
    - iAS: 12
    - Orion: 10
    - Silverstream: 7
    - jRun: 5
    - jBoss: 5
    - Pramati: 0
    - EAServer: 0
    - jOnAS: 0

    Always thought that's a better way to evaluate which basket to put my eggs into. See who's first on the list ? .NET =~ sum of J2EE application servers. And it is not final release yet...

    Where can I download C# specs ? :) Nah, I'll give another chance to J2EE.

                    Yann
  8. I'll put my eggs into J2EE. Type J2EE on www.jobserve.com and you get -> 551. Seems like that basket is a little bigger Yann.
  9. That's strange. Monster gives much different numbers:

    - .NET: 316
    - Weblogic: 484
    - Websphere: 579
    - J2EE 796

    Another MS evangelist posting?
  10. "That's strange. Monster gives much different numbers:

    - .NET: 316
    - Weblogic: 484
    - Websphere: 579
    - J2EE 796 "

    These numbers are about right. If you conservatively figure that it takes 3 times as many developers to develop on websphere or .Net as it does on Weblogic.


    That also could explain why IBM services revenue is so large....
    Matt
  11. If developing servlets, jsps, ejb, etc... are based on a standard and therefore the code is the same, then how can it take 3 times as long to develop in WebSphere than WebLogic, considering WebLogic didn't even have their own development environment until recently, you had to create deployment descriptors by hand (including Entity Bean to Datbase mapping), and package the app yourself. (Writing ANT scripts to do this can is not specific to WebLogic)

    If you mean the act of deploying, sure you can plop the ear and test, but WebSphere development environments (WSAD) accomplishes the same thing woith one click.

    If you are comparing development time, then this should be an attack on IDE's rather than App Servers based on the same standards.
  12. "Another MS evangelist posting?"

    You talking to me ?? I've got friends in Sicily you know... :)

    Well, it's actually quite the contrary. It's not that I'm particularly religious about it - I think anyone could just implement good solutions using good practices with any technology, any language, any IDE - but I'd rather implement good solutions using a technology that does not benefit companies with monopolistic practices which I strongly disapprove; .NET is certainly not as bad as non-MS people tout but as long as J2EE allows me to do the job quickly and cleanly to my customer's benefit, as long as I feel there is this huge community and momentum behind, I'll stick to it.

    I usually check on Jobserve because you get more contracts there. Monster is more about permanent jobs. If Monster gives better J2EE matches, all the better! I'm actually very much in favour of J2EE. Damned, I hope I didn't spend all that time learning a technology that is doomed! :) I think .NET is a reality, that many people are going to like it but I want J2EE to stay ahead of it for ethical reasons, even if Sun is not the best as far as ethics are concerned. At least, it is an alternative.

    To me, .NET is a serious danger that should not be overlooked or laughed at. The fact that there are so many .NET hits on Jobserve or Monster is a threat and we should not hide away from it. At my small scale, I use my influence to make client companies choose J2EE over .NET and I don't really care which application server they choose as long as I know it can fulfil their requirements. On the other hand, J2EE companies should be aware that if they want to win the fight over Microsoft, they should provide us - little Java soldiers - with better weapons. BEA with its Workshop is certainly a great start.

    So don't you ever call me MS evangelist again! Dig it ? ;-)

                    Yann
  13. "At my small scale, I use my influence to make client companies choose J2EE over .NET and I don't really care which application server they choose as long as I know it can fulfil their requirements."

    Well, I actually do care a bit. Some products are IMO *much* better than others. I'm not going to disclose my own preferences but recommending a bad J2EE product will most certainly bring new clients to Microsoft who will hardly be able to back off due to lock-in.

                    Yann
  14. I'm not quaking with fear yet at .NET, despite the number of jobs on <fill-in-the-blank> job board. In 17 years of IT, no company I have worked for has gone all microsoft - and in fact didn't ever get much beyond email servers and maybe sql server. I know there are such shops (all microsoft), and some high-profile ones, but it is going to take a lot more proof that an all MS platform is ready to handle mission critical applications to convince many companies. J2EE is far from being run into the ground - it is growing and improving quite nicely, despite the bickering that happens amongst the various interested parties. Obviously Microsoft doesn't bicker with itself!

    I don't think that if a company has a bad j2ee experience and they are running on Sparcs or HPs or Linux or even Win 2K servers, they are going to run to Microsoft to fix their problems.

    J2EE products may not have a nifty-neat-o drag-n-drool tool like VB.NET, but as the technology and specification matures, productivity tools become more widely available and much more functional. Companies have a wide array of price and feature options in the J2EE world, and if they have a bad j2ee experience with one vendor, there are a lot of others to go to - all of whom are interested in their business. Microsoft doesn't offer that option and I doubt it ever will. If you have a bad experience with them, well - what do you have? For most of my clients, .NET isn't even on the radar.

    I see strong growth in the J2EE app server market both in terms of products, value-adds and jobs, despite .NET.



  15. You're absolutely right. However, so far, Microsoft has never had any real enterprise scale application server. This multiple-language feature might be tempting for some and they intend to port it to other platforms. It's still way behind a stable multi-platform release, it's even way behind a stable release on Windows (wait for SP3). But I've seen that scenario happen already, for a big big company that has now turned to .NET.

                    Yann
  16. Hi All-

    As a BEA spokesperson, I wanted to make some comments about this market study and our market share.

    First, BEA is still the unanimous #1 market share leader -- we either lead or share the lead in market share for all of the major market assessments.

    Second, and this is the big point, Giga's survey is unique because it is the only study that takes into account services revenue as part of the marketshare percentages. In contrast, Gartner Dataquest's pure license revenue model gives BEA a 10 point lead over IBM.

    Third, IBM's growth is most likely in services, not licenses. In contrast, our figures are almost purely license-based AND we aren't allowed to factor in the $2.1 billion in services revenue due to BEA projects done by our SI partners.

    Fourth, we have stronger momentum than IBM right now. We closed 700 *new* customers last quarter alone and we can point to tens of thousands of live deployments.

    Fifth, it's likely that a more probably market share assessment is based upon live deployments. META Group's study is not based upon license revenues, but upon live deployments. Their study for 2001 deployments puts BEA with 37% share and IBM 22% share. For EJB-based deployments BEA is at 52% and IBM is at 14%. These are numbers that are probably much more relevant to J2EE developers.

    Thanks!
    Tyler Jewell
    Director, Technical Evangelism
    BEA Systems
  17. Quote - "we aren't allowed to factor in the $2.1 billion in services revenue due to BEA projects done by our SI partners. "

    Neither was anyone else, right?

    Quote - "META Group's study is not based upon license revenues, but upon live deployments. Their study for 2001 deployments puts BEA with 37% share and IBM 22% share. For EJB-based deployments BEA is at 52% and IBM is at 14%. "

    Wasn't that study commissioned by (and paid for by) BEA?
  18. Tom, for the services revenue, Giga actually does factor in the revenue garnered from IBM Global Services, but they do not factor in revenue from our top tier services partners. So, there is a discrepancy there.

    Tyler
  19. Tyler from BEA-
    Last I checked IBM WebSphere has a significant partner channel and BEA has a significant professional services operation.

    Also if my memory serves me right BEA only began embracing partners in 2001.

    But now for the $2.1 billion dollar question:
    If your "partners" had 2.1 billion in services revenue and let's be generous in estimating that they produced 210 milllion in license revenue- are you saying that a WebLogic project has a 10 to 1 services to license cost ratio? Wow your product must be really, really hard to use! And any shop looking into WebLogic ought to be very wary as to 1) why SIs are pushing it and 2) how much that project will actually cost.

    I look forward to your reply.

    Ken
  20. <Ken>
    ...
    blahblah
    ...
    But now for the $2.1 billion dollar question:
    If your "partners" had 2.1 billion in services revenue and let's be generous in estimating that they produced 210 milllion in license revenue- are you saying that a WebLogic project has a 10 to 1 services to license cost ratio? Wow your product must be really, really hard to use!
    ...
    blahblah
    ...
    </Ken>

    Aha, what a charmingly naive conclusion! I presume, the SIs like Accentures get paid big bucks to only troubleshoot Weblogic applications! Well, the reality is, these people are called in to architect/design/implement applications which also happen to use Weblogic as an application server. Of course, because of the tie-ups, these SIs may well recommend weblogic as the application server of choice, but IBM would do the same too with their SIs. Bottomline is, including services revenue figures in a market share study somewhat skews the real-world deployment numbers.

    --Dave
  21. Dave or is it Dan-
    I was merely pointing out that the 2.1 Billion figure IS ridiculous to lump into the market share figure. The entire market is listed at 2.19 Billion!!!

    He did make the assertion that BEA projects generated 2.1 Billion for Partners in services. I understand what services consist of, thanks for the refresher. I also understand that most projects where staffing is outsourced have at least a 5 to 1 ratio of services to license. But the claim that BEA Partners had 2.19 billion in services related to these types of projects is a bit absurd.

    That being said WebLogic is pretty strong, very stable product.

    An additional question to ask would be what percent of the partner business was done on Tuxedo/WL projects and does BEA separate revenue for WebLogic deals where Tuxedo is a piece of the pie? If not that skews there license mix as well. Same for IBM and WS "portal" deals.
  22. 5 to 1 services/license is considered average.
  23. Of the 1 Billion in revenue that BEA garnered last year, 650 million was in WebLogic licenses. 350 million was earned in support and services -- I don't know the break down of which is what, but our model last year was not to take on major services projects, but to defer all major projects to SIs. Given our limited services revenue and 650 million in licenses and teh 5:1 ratio average, 2.1 Billion for SIs seems reasonable.

    All we are saying is that a big portion of revenue generated due to WebLogic sales are external services companies that aren't factored into this study. You should just keep that in mind when looking at these figures.

    Tyler
  24. <quote>
    You're absolutely right. However, so far, Microsoft has never had any real enterprise scale application server. This multiple-language feature might be tempting for some and they intend to port it to other platforms. It's still way behind a stable multi-platform release, it's even way behind a stable release on Windows (wait for SP3). But I've seen that scenario happen already, for a big big company that has now turned to .NET.
    </quote>

    Oh, I have no doubt that .NET will capture market share - there's nothing like competition to shake things up a bit - but J2EE vendors aren't going to sit around and wait to be swallowed up, either. It will be interesting to see how this all plays out. I wouldn't worry too much about having learned a dying technology, though.

    Cheers
    Ray
  25. Hi Ray,
    >> I don't think that if a company has a bad j2ee experience and they are running on Sparcs or HPs or Linux or even Win 2K servers, they are going to run to Microsoft to fix their problems

    I agree on this & i am pretty sure that everyone knows this. It is not technology. It is something else.. In my opinion they are

    a) Time to market
      - I am sure solutions developed in .NET will be much faster becoz. Lot of things are taken for granted. No vendor selection,evaluation phase etc. Leverage inhouse programmers( C,COBOL etc )

    b) Price
      - I guesss when people evaluate .NET over J2EE solutions they don't consider JBOSS with Linux

    c) Developer friendly tools
      - Developers want to work with nice cool tools right. So developers will support for .NET ( atleast VB developers )

    d) Vast pool of Developers( moderately skilled )
    See the Headlines of this link

    I see these are the real threats. We shouldn't allow what happened to CORBA. Better technology doesn't mean ,it can sustain by itself !!!

    What are we going to do from J2EE side to overcome this...

    One Suggestion to Floyd & theServerSide team:
    ---------------------------------------------
    The entire community in this site is focussed only AppServers,Practices etc. But i guess we should also include "Serverside" tools into this discussion. We had some informal threads for the J2EE Serverside tools but that is not enough. The "REVIEWS" section of this site,should have "reviews" of each plausible J2EE serverside tools.If not everything atleast it should have
    - IDEs
    - O/R Mapping tools

    By including this in this site,it will will help us to effectively compare .NET with J2EE Application stacks.


    Thanks

  26. Hi Murali -

    I would counter:

    <There is no guarantee that .NET will give you faster time-to-market. In any company that I have been involved with, there would still be an evaluation phase of deciding to use .NET. To do that, you need to evaluate your existing infrastructure to determine if it would even be a cost effective move. Maybe a lot of your programmers already know Java - its being taught more and more in college programs these days. Or maybe they're like me - originally a C programmer, but would rather use Java - while it is nifty to be able to use whatever language, for me it is not a big incentive (just my personal opinion).

    <With the variety of application servers available that run on a wide array of hardware, I am sure that anyone can find a J2EE solution to rival any Microsoft solution in total cost of ownership.

    <
    That's up to the opinion of developers - some like Java and other languages, some like VB - there are some very cool tools in the j2ee world. And with time, they will mature quite nicely. (And of course VB developers will support .NET).

    <
    Its not nearly as difficult to find moderately skilled Java developers as it once was, so I don't see it as a major selling point.


    I think .NET is going to be an alternative to J2EE but I don't see it as J2EE's final curtain. I think the J2EE vendors and community are strong enough to put up a spirited fight in the market place and to maintain a strong market share.


    Cheers
    Ray
  27. <quote>
    - .NET: 316
    - Weblogic: 484
    - Websphere: 579
    - J2EE 796 "

    These numbers are about right. If you conservatively figure that it takes 3 times as many developers to develop on websphere or .Net as it does on Weblogic.
    </quote>

    Matt, thats one way to look at it, if you are intent on shying away from the reality :)

    Have a good one
  28. Hi,

    don't be afraid of finding so many .NET-hits on monster or jobserve.
    Those numbers are not comparable to J2EE hits because every headhunter who is looking for somebody with MS skills automatically includes .NET as well.

    In most advertisements .NET is only a metaphor for MS technologies like VB, VC++ or other. Only a few are really working with .NET...but a lot with J2EE...

    Oliver dot Lauer at epost dot de

  29. remember, websphere is more than just the app server. it's a brand -- includes MQSeries and a bunch of other stuff.

    whatever.. remember, these industry analysts make their money by 1) being paid by customers as consultants and 2) by contracts with vendors. take it all with a grain of salt.
  30. J2EE market shares revealed for 2001[ Go to top ]

    The only app server players that will be in this market in 3 years are:
                      Oracle
                      Microsoft
                      HP
                      Sybase
                      BEA
                      IBM
                      Sun
                      Borland (Maybe)

    Could you way why you left out JBoss? It seems to me that Jboss fullfills your criteria for being on the list.
  31. "Could you way why you left out JBoss? It seems to me that Jboss fullfills your criteria for being on the list."

    Because JBoss is in a class by itself. :-)

    -- jason

  32. J2EE market shares revealed for 2001[ Go to top ]

    Why not JBoss?

    I don't know. I haven't seen much about their activities, but from my quick perusal of their webpage they look very interesting. The question really is will ISV's build the value added services I mentioned to run in JBoss or will JBoss (or the open source community) fill in these features in the future. That is the key question and the answer is maybe. Just as many ISVs provide applications for Linux and make their product work with Apache, so may be the case for JBoss. Time will tell. But consulting and support sales alone is a tough business model. I like the idea of Open Source because it allows developers to fix bugs themselves and not be at the mercy of tech support. But without the ISV support this App/Server may have a long life as a departmental app server.
  33. How do you measure market share?[ Go to top ]

    Looking more carefully at the article it seems that market share is defined by the share of the market. ie. how much money has been spent percentagewise across the various app servers.

    jBoss would never appear since it is open-source - ie. free

    Moreover, how do you measure this market share? IBM is in the habit of giving away licenses with hardware. These may never be used, but they can count this license cost and increase their market share. Oracle are moving 9ias into their 11i range of apps, does this equal a license? Sun may bundle iPlanet with their OS - is this a license? Do BEA count the exorbitant cost of portal and integration in their market share?

    What would be really useful is an unbiased comparison of actual deployments, in production, per application server. I think this is what the Oracle Hurwitz survey was getting at. I would be really interested to see how jBoss fares. They have hundreds of thousands of installations, some percentage must be in deployment.

    If anyone knows of such a survey, please post it here. Otherwise, maybe serverside should start a poll...

    Dave
  34. How do you measure market share?[ Go to top ]

    A good way to evaluate market shares is to start evaluating market: a market is traditionally expressed in revenue or people/companies. The people/companies view reflects product penetration and gives an idea about how hard it is going to be to get extra customers. The revenue view reflects the health of the companies. In short, a company which has a high people/companies market share with a small revenue market share is in trouble because it has few hopes of getting more clients and therefore of increasing its revenue market share. On the other hand, a company with a small people/companies market share and a high revenue market share has a brighter future. In short, if you consider Giga Information Group's report as a revenue market survey and the Hurwitz's one as a penetration survey, you will conclude that the fact that Oracle has a 37% penetration market share for a 6% revenue market share is probably the sign of a bad marketing strategy. BEA and IBM seem in a much better health to me. However, keep in mind that those figures from the Hurwitz survey are probably completely wrong: iAS comes bundled with other Oracle tools such as Form Server, which is the only way to bring Oracle Forms to the web. As you probably know, Oracle Forms used to be quite successful and many clients web-enabling their legacy systems bought the Oracle package for that sole purpose. Should that be accounted in the market shares ? According to Hurwitz survey, yes.

    Of course, JBoss is not applicable here. You'd rather compare market shares based on revenue from services. I don't know if such a survey exists but knowing that services are the only source of revenue for JBoss group, it should be fairly easy to know how good they're doing overall.
  35. Looks to me that you are having some problem with Oracle ?
  36. "Looks to me that you are having some problem with Oracle ?"

    I have no problem with iAS, only with Oracle hype. I honestly hope that they get a nicer product for their next release. But I just can't help being a bit sarcastic after all their hype. Go to their home page (http://www.oracle.com). There is so much bullcrap there that I'm amazed as to how people can still trust them. Other vendors do the same but reasonably. :)

    However, I agree, BEA should update its homepage (and its http://www.bea.com/press/releases/2002/0221_ecperf.shtml page) to consider recent ECPerf results.

                    Yann
  37. How do you measure market share?[ Go to top ]

    Dave: "IBM is in the habit of giving away licenses with hardware."

    Not exactly. You can buy the hardware without the software for X dollars, or the hardware for 75% of X dollars with the software bundled for an additional 20% of X dollars. For some reason, their customers always buy the software.

    When it comes to J2EE deployments in the field, we see mostly BEA Weblogic. Probably 2/3 or more. That isn't necessarily a scientific survey ;-)

    Peace,

    Cameron Purdy
    Tangosol, Inc.
  38. Hi,

    I think it is not so important who is leading the market. We need IBM and BEA to push J2EE and it is good that they are opponents in the market. This is the positive thing about J2EE and this will let J2EE win against .NET. Because competition always pushes technologies.

    I am sure that in some years the only app servers will be WebSphere, WebLogic, Oracle and JBoss. But from my experience (in Germany) WebSphere and Weblogic are the only players in the market. The advantage of WebSphere is IBM Global Services and their contacts to the customers. And don't undervalue the power of S/390-zSeries, where WebSphere is the undisputeable number 1. I know a lot of customers (mostly banks and insurance companies) who have so much know how on the host that they don't want to change to a Unix system - particularly not if you need reliability and real security.

    So in my opinion BEA is leading the Windows/Unix market and has the best technology and IBM takes its power from the mainframe and Global Services.

    By the way, I read in a German news paper, that IBM invests more money in WebSphere research (over 1 bill $) this year than BEA has overall revenue...let's see if BEA can defend its position.

    Mirko

  39. <quote>
    don't undervalue the power of S/390-zSeries, where WebSphere is the undisputeable number 1
    </quote>

    Why do you say this (why Websphere "undisputed No.1")?


  40. I said this, because I have never heard of an running appserver on OS/390 oder z/OS that was not WebSphere/390. Maybe on Linux/390 there are other implementations. I read on the BEA website that they have a OS/390 implementation, but I have neither tested it, nor have I seen a running Weblogic on OS/390.

    I thought once, that every appserver implemented in Java would run on OS/390 ("write once, run everywhere") but this is not true on OS/390 (ASCII/EBCDIC problem, etc). I had even problems to get APIs like Apache FOP running (only with a lot affort and changes in the sources we made it run)and you have to use special implementation for XML/XSL parsers (see www.s390.ibm.com/xml).

    Please tell me if somebody knows about running appservers others than WebSphere (on OS/390 or z/os). It would be interesting to know about it, because WebSphere is really not the silver bullet on OS/390 ;-)

    Mirko


  41. I know that BEA have Weblogic running on OS/390 / z/390 in a number of small and large sites. (you will also note that it has been supported in OS/390 since v 4.51)

    You should speak to BEA to get more info.

    Moreover, IONA iPortal is also supported on OS/390.

    It is a bit of a myth that only Websphere runs on IBM platforms...

  42. If they have supported the mainframe since 4.51 and there are projects running with Weblogic on OS/390, how can I understand this press release?

    http://www.bea.com/press/releases/2002/0225_ibm_mainframe.shtml



  43. I would say that it is BEA marketing spin. My guess is that it relates to the 6.1 platform supporting z/390. Probably, its also in response to this.

    BEA's platform support page is : http://e-docs.bea.com/wls/platforms/index.html#os390

  44. The difference between the Mainframe implementation of IBM WebSphere products and the other Application Servers is most of the other implementations of an Application Server are just a port from the distributed platforms. The URL you referred to even claims this being that WLS is a pure Java implementation.

    However, WebSphere Mainframe version and the distributed version are different products, with different administration tools and architecutre. You can run the distributed version on a LINUX partition on the Mainframe, but it is not the same as WebSphere for OS/390 or zOS (It is not ported, but developed on the Mainframe from the ground up). The Mainframe version of WebSphere is developed using native Mainframe libraries to take advantage of Mainframe processors. What advantage is it to run an Application Server on a Mainframe Operationg System if the application still needs to treat the rest of the Mainframe as if it is on a different node.

    If the idea is to treat a partition of the Mainframe as a distributed platform for easier hardware maintainability, then it doesn't matter what Application Server you use. However, but if you want to develop a Session Bean that is an actual CICS transaction, or have special Entity Beans for IMS databases, well WebSphere is the way to go.

    (Views are my own, not necessarily IBM's)

  45. >> develop a Session Bean that is an actual CICS transaction

    What do you mean?

  46. CICS & EJBs[ Go to top ]

    What I believe he means is that you can develop a session EJB that is a wrapper for a CICS transaction. While you could also do this when the EJB server is on a distributed platform, Websphere for zOS can do this with the highest eficiency/shortest code-path. Running the EJB server on Linux/390, or under OS/390 Unix Services has some advantages, but doesn't give the same efficiency, and doesn't take advantage of other OS/390-zOS features such as user/group based prioritized task dispatching from Web Server to App Server to CICS, SMF recording, Parallel Sysplex enablement and others.
    If you are really considering this, you should also look into the latest CICS enhancements. With newer releases, CICS itself can act as an EJB container, allowing transactions to be exposed as session EJBs.
  47. CICS & EJBs[ Go to top ]

    Tom,

    I think you are right. If you have a lot of transactions and need short lock times on the database, maybe WebSphere for OS/390 is the only possibility...

    What I never understood was the WebSphere startegy of IBM. First they tried to bring the Component Broker as WebSphere Enterprise Edition which I experienced, never worked. Now they have a different product, but it seems poor for servlets and JSPs but good for EJBs...(servelts are session beans at runtime on WebSphere/390, as I know). So I think, that they want WebSphere to be a kind of EJB only container as their third transaction monitor on OS/390. (By the way, I think that OS/390 is too expensive and slow for Servlets and JSPs)

    I would have taken CICS, which has a very good reputation, add EJB features and thats it...

  48. CICS & EJBs[ Go to top ]


    ".. you can develop a session EJB that is a wrapper for a CICS transaction"
    I still dont understand what this actually means. What are we wrapping?


    It is not clear to me how such a "huge" performance gain can be achieved while still respecting the EJB and J2EE standards. I smell a proprietary EJB - but I would be happy to be proved wrong.


    It is also not clear to me that Websphere for OS/390 has, in fact, achieved this huge performance gain. Certainly none of the documentation I have managed to find suggests a revolutionary new architecture. Furthermore, there are no published performance figures comparing even Websphere4Z/OS and regular Websphere. (I would be happy if the IBM guys here could point me towards the right information - it would be really helpful for me at the moment).


    *If* it were the case that Websphere for Z/390 were fundamentally different ("developed on the Mainframe from the ground up" to quote an IBM emloyee), then I would have some serious doubts about how well the *two* servers would be simultaneously maintained. The dual development cost would be large and there would be asymmetric advancement of each container. Both would take longer to release - or one would become the poor-cousin in terms of features and standards support (and released even later than normal). (Its for this reason that BEA and IBM discontinued further development on their respective Tuxedo and CICs EJB containers.)

    In any case, the entire selling point of WAS4OS/390 is based on the assumption that OS/390 is a killer server platform. While there are lots of MVS mainframes that everyone is very keen to leverage/integrate/support, is this perhaps an effort by big-blue to push its big-iron? Maybe, maybe not - maybe I am too cynical.


    (Views are my own, not at all IBM's ;-)
  49. CICS & EJBs[ Go to top ]

    "is this perhaps an effort by big-blue to push its big-iron? Maybe, maybe not - maybe I am too cynical."

    I agree with you Nick. This is clearly a sneaky attempt by IBM to sell 390 hardware/software by adding new features. They even have the nerve to make it J2EE compliant!
    I suspect that it is all part of a grand plan to make a profit.
  50. CICS & EJBs[ Go to top ]


    Yes, you're right, that was unnecessary...
    However, I work in an IBM shop. I have to live with the results of their sneakiness... ;-)

    (BTW: is it compliant?)

  51. J2EE market shares revealed for 2001[ Go to top ]

    BEA and IBM share the first position. I was quite surprised to read that as I heard from different sources that IBM was now before BEA after a tremendous upwards momentum. Another source though stated that BEA was at 37%, IBM at 22% and Oracle at 11%: http://serverwatch.internet.com/news/2001_09_26_a.html.

    Apparently, IBM and Oracle have been giving hard discounts on their application servers to better penetrate the market. That explains part of IBM's success. Oracle iAS did not meet the same success as it seems that they turned Orion into a buggy product (second hand but reliable information) and clients are not so keen at using it. They will certainly improve for the best of J2EE but here we're talking about 2001 results.

    One should always take those market surveys with a pinch of salt anyway and not get too emotional about it. Just like benchmark results.

                    Yann
  52. <quote>
    Oracle iAS did not meet the same success as it seems that they turned Orion into a buggy product (second hand but reliable information) and clients are not so keen at using it
    </quote>

    From personal experience, the j2ee container itself (oc4J) isn't that buggy - but the whole 9iAS, while a nice idea, does need a bit of work.

    Cheers
    Ray
  53. What a shocker!

    Predictions for 2002:

    I predict that BEA and IBM will lose some share to Oracle this year, with Oracle 9iAS easily putting iPlanet in the dust. I predict that there will be more of a commoditization of the app server market and that smallish to mid range companies may go against the Giga advice and choose app servers that aren't on the Giga "approved" list in the article as budgets will still be a consideration. Those app servers will be Orion, Pramati and JBoss.

    Just predictions, understand.

    Cheers
    Ray
  54. If anyone knows the application of John Forbes Nash's (go see "A Beautiful Mind", great flick) theories on groups and behaviour prediction then perhaps they could apply this to the ASSP market and be at least as accurate in predicting the outcomes as anyone else... all opinions respected :-).

    Most of the analysts seem to think that the market will shake out to 3-4 players in 3-4 years. Those would naturally be the big guys:

    Oracle,
    IBM,
    Microsoft, and;
    perhaps one other

    In case you were wondering, Gartner has already stated that BEA needs to seriously re-invent themselves to stay in the game long term.

    Maybe jBoss will become the Linux of the J2EE world. It needs to evolve and incorporate the solution areas that the other players are now taking on. Order of integration complexity n-squared :-0


  55. 650 Million in WebLogic Licenses and 350 in Services-
    What about Tuxedo? That shouldn't be considered J2EE App Server Market. I think that's a fairly large market. Plus BEA has a host of other middleware products and even some applications. So the true number is skewed no matter how you look at it. But at the end of the day BEA is the most Popular COMMERCIAL Application Server (most paid deployments), IBM is the best company in the world at selling (most discrete revenue) to it's base (take note Oracle I said selling, not extorting) and by the time it's all done both companies will make a ton of dough and Bill Gate's will still be rich and ugly.
  56. J2EE market shares revealed for 2001[ Go to top ]

    As my Maths teacher was found of saying "I think this discussion shows that you can prove anything with statistics".

    The market leaders are by far IBM and BEA Weblogic. It is great that these two companies are fighting it out for the App Server Crown. They will of course need to keep an eye in the rear view mirror to make sure some company does not come in and steal the crown from the both of them.

    Competition is a great thing for us programmers and surveys like this and performance results encourage it. Better this than a MS world. You go guys!

    David
  57. David,

    "Most of the analysts seem to think that the market will shake out to 3-4 players in 3-4 years. Those would naturally be the big guys: Oracle, IBM, Microsoft, and;
    perhaps one other"

    Most, eh? Nice straw man. Most of the analysts seem to think that BEA has a huge lead in the app server space, that Oracle is just managing to hang in there (if they are present at all), that IBM is leveraging everything but the kitchen sink to get any numbers whatsoever, and that Microsoft is in their own little world. That's my straw man, anyway ;-)

    Peace,

    Cameron Purdy
    Tangosol, Inc.
  58. I agree about the current BEA lead for sure. I would be crazy not to so I am not disagreeing with you. The longer term is hard to predict but at the end of the day we are talking about $25b+ companies vs $1b+ companies. In the longer term, unless there are strong differentiators, the small guys will suffer from the sheer S&M clout of the big guys.

    For example, look at the call-to-war that M$ is launching now with .NET. Even Scott McNealy is scared (see his JavaOne speach..."Mankind Needs Your Help").

    M$ have funded the marketing of .NET with $200M+ this year which is about as much as they commit to a new OS launch. My point is really that the small guys are going to have a hard time with the big guys (bogo-market-share, skewed figures or whatever) since all of the big guys have turned thier attention to this market and will be kicking the sh*t out of each other and anyone else in the ring.

    :-)

    David