Discussions

News: BEA Claims Big Wins Against IBM

  1. BEA Claims Big Wins Against IBM (28 messages)

    In a recent press, BEA claimed that more than 210 customers in the Americas, Asia/Pacific and Europe selected BEA over IBM in head-to-head sales situations during the past quarter. 125 of these customers selected BEA despite the fact that IBM WebSphere was the incumbent platform. 93 of these customers chose BEA for new projects, and in 32 cases Websphere was replaced with Weblogic.

    Read BEA's Press Release "BEA Wins Hands Down in Real-World Competition Against IBM as Customers Prefer BEA"

    ******************************
    Editor's Note *
    ******************************
    "Disappointing that TSS publishes this kind of one sided press release. Publishing stories by 'independant' analysts is one thing, at least they are supposed to have no bias but this is pretty one sided, will we see each of BEAs press releases posted on this site also?" -Billy Newport

    I would like to apologize for this lapse in editorial integrity. This news item was posted hastily and was not given a proper editorial narrative. TheServerSide's goal is to bring you objective and essential J2EE news and we will work harder to realize it.

    This issued press release is a claim made by BEA, not an independently reported article. What do people think about these claims?

    Threaded Messages (28)

  2. <quote>
    One customer that replaced IBM with BEA is Chicago-based Abt Electronics. "We needed to redesign the Abt Electronics Web site to provide better security, more automation, and more real-time information, but most importantly, an enhanced and more satisfying buying experience for our customers," said Bill Wolfe, chief information officer at Abt Electronics. "We evaluated BEA, IBM, and others. We chose BEA for its flexibility and scalability, and we expect it to dramatically reduce the effort, time and cost necessary to enhance and maintain our Web site moving forward."
    </quote>

    It sounded like a new purchase, instead of replacement, according the CIO's quote. I really want to hear some WebLogic replacements of WebSphere: why, how, cost, roi, etc.
  3. SAN JOSE, Calif., Aug. 14 /Flack Machine/ - BEA subsequently sent TheServerSide.com it's sincere thanks for reprinting its unabridged marketing drivel with a complete lack of editorial context.

    "I remember when we used to pay for advertising," commented Dave Swickers, Weblogic Marketing Director. "Now remember, kids, free advertising: good, free appservers: bad."
  4. funny numbers[ Go to top ]

    According to this release, in the 210 head-to-head sales that BEA says they won, Websphere was the incumbent in 125. That leaves a maximum of 85 customers where Weblogic was the incumbent. Per that sampling, Websphere must have started the quarter with 50 percent more market share than BEA.
  5. funny numbers[ Go to top ]

    This figure doesn't include the others deals that were not head to head. Just as IBM has a number of deals w/ no competition so does BEA.
  6. Disappointing that TSS publishes this kind of one sided press release. Publishing stories by 'independant' analysts is one thing, at least they are supposed to have no bias but this is pretty one sided, will we see each of BEAs press releases posted on this site also?

    If it even had some kind of journalistic narrative besides the comments from your readers then I would see some justification for sticking it as the lead story but frankly I expected better than this.

    Given BEA started off with almost all of the market in hand about 3 years ago and are now struggling to keep their #1 slot, that has to be scaring them silly.

    Here are a couple of things to ponder on when observing the coming 'battles' and to put 210 customers in perspective. Given we have no narrative on this thread, here is my 'spin' on a possible narrative, right or wrong.

    Given the current drive towards lowering costs and as application servers mature and commoditize, vendors such as IBM, Oracle and Microsoft will have a serious advantage over BEA in any contest because they sell application servers, development tools, databases, enterprise messaging, integration platforms and services, operating systems, hosting facilities, management software and hardware. As this stack gets further integrated across these products, license costs and operating costs for these bundles will come down. As application servers become more commoditized this will become more and more important as a deciding factor for customers when choosing a platform provider (notice, not an application server, I said 'a platform').

    The next point is easy to forget for us developers, so I'll state it although it should be obvious.

      "The only companies that care about technology for it's
       own sake are technology companies".

    For everyone else, technology is a necessary 'evil' to do business efficiently and make no mistake, they are looking to get the best return on every dollar spent on IT.

    Many corporations are starting to consolidate multiple J2EE applications and things like databases around the firm on to 'standardized' server farms in order to lower costs and reduce the number of licenses required. Work out how much a HA setup for a typical application costs (2 app server boxes, 2 boxes with shared DASD for a database, HA veritas etc software, now add in software licenses, SAs, backup costs, operating costs) and you can see why it's a question of when not if this will happen.

    Once they do this, then a good chunk will figure out that it's cheaper again to outsource this to a third party hosting company, further reducing license revenue as license are shared among a larger pool of applications. As the platform matures, they'll upgrade less, which will suit the hosting companies as it's less work for them and it suits the customers as it's less work for them too. Again, license revenue drops.

    Even smaller companies will do this once the big guys create the market and lower the cost of entry for this service. More consolidation of license revenue occurs and it drops again.

    Enterprise software companies will be probably be forced to switch to a rental or capacity used model of pricing for revenue from middleware and applications to deal with this consolidation of licenses and the increasing power of the CPUs those license are deployed on (which also reducing per processor revenue especially in shared environments).

    IBM calls it the fifth utility or eHosting. Once this starts to happen widely and it's gonna, a company whose bottom line depends on license revenue from commodity software which is only a small piece of the puzzle is going to be in jeopardy and it's going to get squeezed, and squeezed hard. Microsoft is the only company that have survived this commodization in the past and they had a defacto monopoly to alleviate pricing pressure thus avoiding the trap. BEA does not have a monopoly today and isn't likely to have it moving forward either.

    Look at what Oracle, SAP, Intel, IBM, GE etc are investing in. It's costing billions to make the investments needed to carry on as a player in the market environment which is coming. What are they doing? They are investing in ehosting type infrastructure and it's only for the very deep pocketed. This is where enterprise software is moving, not J2EE v10 or JMS v4, it's becoming a service for customers to use just like they use water or electricity today.

    Now look at BEA and what they are investing in and tell me how important is it that they have 34% versus 32% for IBM right now when looking ahead over several years. But wait, it's troubling isn't it, maybe BEA knows something which GE, IBM, Oracle, Intel and Microsoft et al have missed, are all the other companies stupid and throwing billions away? are Andy Grove, Ballmer, Ellison or Sam P clueless?

    Read between the lines here and see whats going to happen next. Someone is going to buy them and BEA can spin this anyway they want but I think this is inescapable, it's that or fade away.

    Billy
    (Opinions are my own and don't represent IBM)
  7. "The only companies that care about technology for it's own sake are technology companies".

    Billy, i'll save your message in my archives. Sometimes geeks forget what this game is all about. Thank you for the reminder.

    Markus
  8. Billy, i'll save your message in my archives.

    > Sometimes geeks forget what this game is all
    > about. Thank you for the reminder.

    Yeah, me too.. as always - it's a pleasure to read his comments!

    -- Andreas

  9. a message for posterity.

    remember how engineers disliked working for those giant power utilities ...

    all the chaotic fun we have will be over soon
  10. " Enterprise software companies will be probably be forced to switch to a rental or capacity used model of pricing for revenue from middleware and applications to deal with this consolidation of licenses and the increasing power of the CPUs those license are deployed on (which also reducing per processor revenue especially in shared environments). "

    Did you mean the ASP model will again enter into the IT shops of the big enterprise ?
  11. Dherraj,
    Yes, big first, and once cost barriers come down then middle sized and then smaller ones.

    Billy
  12. Billy Newport writes :
    >>Given BEA started off with almost all of the market in >>hand about 3 years ago and are now struggling to keep >>their #1 slot, that has to be scaring them silly.

    This has a lot to do with the fact J2EE is an open standard
    implemented by multiple vendors and is to be expected.

    What's odd is that IBM,HP,SUN,Oracle, et al failed to dislodge them from the nunber 1 slot so far.

    Oracle and SUN will gradually take market share away from BEA and IBM, that is also to be expected.

    Looking at how companies spend their cash, IBM might like to consider investing in the WebSphere app server, not eHosting.

    Will they be the last major vendor with a 1.3 solution ?

    Think of the billions spent by companies paying for IBM's profesionl services to come in and salvage various WebSphere projects.

    This is the first area for saving, and is simpler to tackle then eHosting.

    Oracle's mainstay is their RDBMS offering, It's a good easy to use product, are they a weak, one product company ?
    How about SAP ?

    Do you consider having a product that is easy to develop with, requires minimal external consultancy, and supports the latest standards "Technology for it's own sake" ?


  13. What's odd is that IBM,HP,SUN,Oracle, et al failed to dislodge them from the nunber 1 slot so far.

    <bn>
    Very high by the end of the year.
    </bn>

    Will they be the last major vendor with a 1.3 solution ?
    <bn>
    It's possible and thats not the most important thing on the radar. I've personally heard customers 'complain' that the products are rolling out faster than they can keep up.
    </bn>

    Think of the billions spent by companies paying for IBM's profesionl services to come in and salvage various WebSphere projects.

    This is the first area for saving, and is simpler to tackle then eHosting.
    <bn>
    Right, IBM has and is spending a fortune on tools which allow J2EE enterprise applications to be developed more easily than before. Look at eclipse and WSAD. There are also two costs here, one is development, one is deployment. Tools needs to step up to make development cheaper, deployment will probably get cheaper through hosting.
    </bn>

    Oracle's mainstay is their RDBMS offering, It's a good easy to use product, are they a weak, one product company ?
    <bn>
    Oracle hosting, Oracle Applications suite, ... hardly a one product company. They have bet huge on oracle applications as well hosting facilities so customers can outsource their applications as well as databases etc. They are heavy investors in network computing as well as Larry investing in that Oracle Accountancy eapplication which is available now.
    </bn>

    How about SAP ?
    <bn>
    Building mySAP.com
    </bn>

    Do you consider having a product that is easy to develop with, requires minimal external consultancy, and supports the latest standards "Technology for it's own sake" ?
    <bn>
    Ease of development will soon have little to do with the production application server. Standards will let tools vendors ship lightweight, RAD friendly J2EE servers which are used for development. Deployment can use different servers with different characteristics. Although, the J2EE standard is a ways away from making applications this portable as this point. I think to compete with Visual Studio, we'll see builtin application servers very tightly integrated with tools. So, if you're in the application server business then you need your own tooling or you better be in bed with a tools vendor and working very closely with them.
    </bn>

  14. Billy,

    You often state that your views are your own and not IBM's, however some of what you say here sounds like it comes directly from an IBM marketing handbook. I say this becasue I have heard it before - from other IBM employees (and, more recently, from Sun).

    I am not doubting your technical ability, and a lot of what you write is quite interesting, however, your one-eyed support for your employer, IBM, and taking potshots at competitors - BEA in particular - is not really very professional. It would perhaps be more acceptable if you were "independant analyst", but you are not - and the bias is evident (in this post and others). Why do none of the guys that post here from BEA, Oracle, HP, Borland, Macromedia etc feel it is necessary to lower the level of debate to "we are going to put them out of business" - "they must be scared silly"?
    I know that you were reacting to the bias of the orignal post (now corrected) but it wasnt posted by any of the BEA guys, was it? And moreover, you didnt actually deal with any of the points in the article...


    Before I continue, for the record let me point out I dont work for any of the vendors - never have. Moreover, I dont hold any shares or any interests in any either. I work for what is known as a "customer" - something that a lot of the larger vendors seem to forget now and again. The company I work for is a customer of IBM, BEA, Oracle, Tibco, Microsoft, Sybase and many other vendors, so there is no hidden agenda - no cosy relationship. Currently, on the appserver front, BEA happen to give us what we ask for. As soon as they stop delivering, they will likely be replaced by a vendor who does deliver. If that happens to be Websphere, then great. That is what competition is all about - its what J2EE allows us to do.


    While I dont think the original article that prompted this thread had a great deal of value, I certainly dont think that the resultant debate adds any.

    The original article claimed that some existing Websphere customers preferred Weblogic. I dont plan to argue for that case or refute it - but I will acount our experience as I guess we make up 1 of the 125 customers where Websphere was the encumbent platform (if you believe those numbers that is). (However, I will point out that we are still Websphere customers and will likely continue to use both).

    In the last 12 months or so, I have met a few employees of large organisations, banks etc. It was interesting - a number of them were at various stages of going through a full-pitch internal battle with their "standards" people in order to use (usually Weblogic) over Websphere. It was interesting because at that time, we were also going through our own exercise of getting Weblogic "approved" as an alternative standard to Websphere. Interestingly, there were no cases of Websphere replacing any appserver.
    Many of the complaints of Websphere that I heard during these discussions were the same - and strikingly similar to our own.

    It is a pain to use: To be fair, a lot of them were using WAS3.5. Also, to be fair, most were using VAJ, which they hated. (it just goes to show that integrated tooling is not as important as good tooling).

    It is way behind in standards: While the december 2001 alpha release of WAS5 generated some optimism that IBM had turned the corner on its stance on standards support, 9 months later we are still waiting for a shipping product. (I hear its coming soon?). (More on why I think standards support is important later)

    Integration with non-ibm products is average: While potentially contentious, this assertion has some basis: On wintel, only the IBM JVM is supported (what about Sun, Jrockit?). Full XA integration with messaging was only via MQ. Support for BLOB's worked with DB2, but not with Oracle.

    It was expensive to maintain: Most of the people had full-time IBM consultants on site that were required to keep things running (in addition to their own support staff). Not because it was utterly unreliable - but mainly because when problems did occur, it was difficult to find out how fix/avoid them. Also, just getting up and running for someone new to WAS too often required assistance.

    The support from IBM was poor: Also a contentious point, there is a general feeling that the level of attention they got from IBM was less than that of most other software vendors. I can give examples of our own current ridiculous situation regarding IBM support, however, to be fair, its not all IBM's fault, so I wont.

    Now, I grant that this experience with Websphere comes from a limited group of poeple - I am sure there are people that are perfectly happy with Websphere. I am sure that there are also those that disagree violently with the points above. Please note that I am not arguing that the above is a generalisation, nor am I trying to extrapolate from that - its just an account of our experience/opinions and those of people I spoke to - readers can take it with a grain of salt.


    With particular reference to the points you raised in your response:

    Integrated Product Stack:
    The arguments that seem to be coming from the big vendors seem to be the same: that it is vital to have a completely integrated stack - and also that this is also what customers want. Any vendor that doesnt have the "full-stack" is going to get squeezed out of business.
    To some extent, this goes against the best-of-breed ethos of J2EE. It also goes against past experience, Databases are a commodotized market, yet Oracle have enjoyed a healthy market share - why? Not because they have sold an integrated stack, but because its a good product. There will always be a market for a good product.
    Customers want choice. Customers dont necessarily want the whole stack - they want to be able to integrate the products they choose without having to pay a penalty. Those that dont want to integrate with anything else would probably be happy going with Microsoft.

    Also, the pricing advantages touted of the bundled-stack are quickly evaporated if extra development is required because there are missing features in any one of the products or if it is considerably more difficult to integrate with *existing* third-party products.

    E.g. If I am using Websphere on Wintel, and I hit a bug in the IBM JVM (which we have done), I want to be able to swap it out (without losing my support) until the bug is fixed. If my application is down or delayed in going to production, the total cost of this will dwarf the *total* license cost - let alone the delta saving in bundled licensing.
    We had a situation where a project was delayed 6 months from going into production because of an AIX/JVM bug. There was not much we could do but wait for the patch (there is only one JVM for AIX). We, and I am sure others, want to avoid these situations where we are left with no choice. Thats why, for instance, we choose J2EE over .Net.

    Standards and Application Servers:
    In response to criticisms regarding Webshere's slow uptake of standards, I have heard the following from IBM employees:
    "If you want the latest bit-twiddling API's, then Websphere is not the right choice. Websphere only supports mature standards".
    I have also seen: "its not the most important thing on the radar" and "The only companies that care about technology for it's own sake are technology companies".

    I continually assert that, for most projects, the cost of development is far greater than the license costs of appservers. While the investment in tooling from IBM and other vendors helps, its not going to solve development problems that arise when the target appserver doesnt support the required features (read standards). Supporting the current standards means that we can leverage more of the appserver features (while avoiding vendor-lock) to reduce the amount of development we have to do. The less development we have to do, the cheaper all around it is for us.

    If we develop something an appserver should give us, it can be *very* expensive. We take an initial hit in development and testing. We take a secondary hit when, inevitably, the first implementation is poor, and will require more work. We take a further hit in documenting how to support it. And then we take a continuous hit in maintaining it and supporting it for the life of the application. And all these costs are diverting resources from the efforts of writing and supporting software that actually *makes* us money.

    Having a product that is easy to develop with, requires minimal external consultancy, and supports the latest standards is not "Technology for it's own sake". It gives us (and, I am sure, others) the potential to save a lot of money. More money than bundled licensing and eHosting will.


    On the Topic of eHosting; I have some doubts about the significance of this. I am not denying there is a market, but I am not convinced that it is so significant that it is going to drive all but giants: IBM, Oracle and Microsoft out of business.
    For many organisations, there will be reasons why outsourcing hosting wont happen:
    + Some will be technical reasons: Legacy constraints will prevent it from being possible.
    + Some will be legal reasons: Some organisations have regulatory/legal requirements that would prohibit them from hosting off-site (banks, for example).
    + Some will be strategic reasons: Will the organisation place that level of trust in the hosting company - will they take that level of risk? I am not sure our organisation would.

    If you think, after reading this, that I am rabid supporter of BEA or have any dislike for IBM or any of its employees, then you would be wrong. I give out credit and criticism in equal measure where its due.
    So why do I post this? Its certainly not for fun - I have had my fill of Weblogic/Websphere debates from our own internal discussions over the past 9 months. The reason is because TSS has a large readership and is influential. (I have no doubt that this is why Billy posted his reply).
    If left unchallenged, then the arguments such as posted above start to appear in peoples decision making rationale. It wouldnt be the first time I have heard some material posted on TSS appear in a discussion.

    Anyway, we can see the gloves are coming off in the J2EE marketplace...

    Regards,
    Nick

  15. Nick,
    From my posting:

    "Given we have no narrative on this thread, here is my 'spin' on a possible narrative, right or wrong."

    A possible narrative. I made no shots on WebLogic the product. I didn't say WebSphere is better than WebLogic or vice versa. I didn't recycle old issues with WebLogic 4.x, 5.x or 6.x or the current 7.0. I wasn't saying product X is better than product Y. It was purely 'a possible narrative' on a possible market direction (right or wrong) rather than whose application server is better.

    Latest product standards.
    Yep, WAS 5.0 will be out in Nov 02. It'll support J2EE 1.3 plus the enterprise version has all the web service composition/process choreography stuff, asynchronous beans, business rules, and the rest.
    You spoke of the issues with customers needing to develop their own plumbing. I agree and this is what we're doing to help. I agree that supporting the latest standards is important but I also think supporting new types of application, plumbing and expanding the reach of J2EE based technology is important also. So, it's a balance. There is a bunch of stuff present in the full product which is very useful to customers and isn't part of the J2EE specification at this point. All of the extension stuff is or will be a JSR very soon now so the extension work helps us gain experience with new features and helps improve the J2EE spec through improved JSRs.

    The integrated product stack.
    Integrated does not mean unintegratable or not competitive on it's own. Each product in the stack should be usable on it's own and be competitive on it's own. How-ever, using multiple products together should result in efficiencies and enhance the competitiveness of the overall suite. Thats the point.

    eHosting.
    This is happening even in markets you think it unlikely. This will become more important in larger customers.

    Billy

  16. >> I made no shots on WebLogic the product
    <grin> No, thats true, but instead you took the opportunity to take plenty at BEA. (its not the first time either). I just think that plenty of people take interest in what you write, however I think that even you would admit that you dont hold an impartial position. It would be sad if TSS was hijacked by vendors to be used as a platform for launching their marketing wars...


    >>Yep, WAS 5.0 will be out in Nov 02
    Oh no. I was told August - I jokingly interpreted that as September... ...very well.
    Like I said, it makes it difficult for us to use WAS, when its so far behind.

    >>using multiple products together should result in
    >>efficiencies
    Agreed. I think people object about vendor-lockin when its unreasonably difficult (or impossible) to integrate other products (or they are "not supported").
    In general, I am wary of companies that sell a "full stack". Its usually in their interest to make it more difficult to integrate products other than their own. On the other hand, vendors that specialise tend to make damn sure that they are integrated with as much as they can - its in their interest to.

    >>This will become more important in larger customers
    Given how conservative larger organisations tend to be, I would have thought these would have been the last to take up out-hosting (for the last reason I outlined). Do you mean larger *existing* customers - ie full-on IBM/Microsoft/Oracle shops?

    Cheers,
    -Nick
  17. Nick,

    In addition to Billy's comments, I also wanted to thank you for taking the time to write such a thoughtful post.

    In general, I hardly think the contents of a BEA-written press release saying "we're better than IBM" qualifies as NEWS. I was surprised to see it posted here in the first place.

    Randy Schnier
    WebSphere Development (colleague of Billy)
  18. I think the BEA sales pitch was worth posting for the quality of response it got from such a talented bunch of individuals.

  19. >>In general, I hardly think the contents of a BEA-written
    >>press release saying "we're better than IBM" qualifies as
    >>NEWS. I was surprised to see it posted here in the first
    >>place.

    Agreed.

    This site is valued for its content and the quality of contribution from readers. As I mentioned above, it would be rather unfortunate if we have to start running everything through the marketing-bullshit-filter.

    -Nick
  20. << Disappointing that TSS publishes this kind of one sided press release. Publishing stories by 'independant' analysts is one thing, at least they are supposed to have no bias
    >> Agreed. However, the only independent analyst I knew personally is working for food. Every one else is getting paid to deliver white papers.

    <  "The only companies that care about technology for it's
       own sake are technology companies".
    >> Or developers who have to make it work. 100% agreement there.
    Chief Fingering Officer to Chief Incompetent Officer: "What is the problem with your developers? Didn't we got a great deal on these licenses from Incumbent Baggage Machines®? Now you are telling me that we have pay these consultants from Imperial Globetrotting Consulting Group® for another year? Plane tickets, meals and all?
    Chief Incompetent Officer, "Yes Sir, it's all the developers' fault. For months, I have heard nothing but whining about lack of features like hot deployment, many-to-many CMR, local ejb and workstations only fit for word-processing. Why, when I was managing that mother of all RPG projects at the Post Office, all we need are flags and dumb terminals.
    Chief Fingering Officer: "Fire the smart-asses and outsource the whole project!"
    Chief Incompetent Officer: "Yes Sir! Consider it done! By the way, we did get a ton of free DataBug2 licenses. Eat your heart out, Larry"

    <>>&#8220;No high-powered executive ever gets fired for buying Incumbent Baggage Machines&#8221;®
    Even the guys from Pee-Wee Consulting(PwC), under new management, can tell you that.

    <>>Carly, it's going to be cheaper than Bluestone(RIP). Get your checkbook ready.

        How true. BEA is losing the big marketing/financial war even while winning many small technical battles. For large enterprises with legacy systems and looking for complete solutions, BEA comes up empty.

    Java Enterprise Developer Incognito

    Disclaimer :
    I have never been a high-priced consultant, high-powered financial executive nor an ignorant IT executive bent on covering my own ass at a Fortune 500 corporation. As such, I make no representation as to having any qualifications to make sound, informed opinions on Enterprise Computing Solutions and the budgeting thereof. In fact, I have none at all. My only qualification is that I get paid to deliver J2EE software that works. It is also the reason I am still making a living.

    Opinions are my strictly my own and don't represent Incumbent Baggage Machines® or Being Executive-decisioned Alive®. I have never been an independent analyst, dependent analyst, employee, H1B-consultant, shareholder, stock-option holder, bean counter of either entities, real or imagined)
  21. "Once they do this, then a good chunk will figure out that it's cheaper again to outsource this to a third party hosting company, further reducing license revenue as license are shared among a larger pool of applications"

    Just tyring to understand the feasibilty of above made statement.Could you exlpain how a JVM could be shared with other xyz org's apps.How can you consolidate apps even with your own/other enterprise.I think you end up spending more in designing,paying to consultants,modifying the apps to reduce license spending.

    How clustering,security,load balancing would work if you mix up things.I see this more a academic approach or lacking some knowledge in this.

    P.S We can also take this offline.
  22. Send me an email at bnewport at yahoo dot com and we can discuss it if you'd like.

    Billy
  23. Hello,

    I don't understand how it is possible for IBM to claim such an important market share. How all these people are doing for developping on a App Server like websphere 5 (I do not know about earlier versions) ? That's for now, a mystery for me. If someone can bring up the light for me....
    Let me explain a little further. I'm working for an independant software vendor and developping the same product on Jboss 3, Weblogic 7, Websphere 5.
    My application is made of :
    1 ear containing 3 ejb-jar + java utils classes :
    ejb-jar 1 : 1 session, 1 entity bean (cmp)
    ejb-jar 2 : 1 session, 6 entity bean (cmp)
    ejb-jar 3 : 2 entity bean (cmp with cmr)
    war : 1 servlet
    + ejb-client-jars
    Total size of the ear : about 1MB.
    My developpement server is Jboss which is perfect (light, fast, easy to use, reliable). Then, when I reach a "stable version", I port it to weblogic and websphere. Weblogic porting and debugging is usually a matter of hours (about half a day), weblogic enforces usually specs compliance a little bit than JBoss. Though it is not perfect (bugs, cmp-jdbc mapping confusing etc...) it is USABLE.
    Now we come to websphere, I'll list below what I was needed to do (maybe I'm missing some things here, but if there are any WAS experienced developpers that point something wrong, I will be happy to know what... because I do not found answers anywhere!!).
    1. Build it
    1.1 Build your ear the "classic" way, (deployable as it is on other servers) (15 secs).
    1.2 Make ant tasks working (WEBSPHERE SPECIFIC):
    - the trick here is that websphere comes with his own "customized" jre (see .properties and dll and so on... under WAS_HOME/java). You must use this jre even for building, if not... you're in big trouble. There is even a .bat to run ant tasks in bin directory !!! The good news is that if ther is no such script, I would not have been able to make it run because there is no documentation (AFAIK) on this !!!
    - the ant tasks documentation is extraordinary incomplete and inconsistent (there are typos, that prove that IBML people don't even reread the doc). Am I the only one on earth that build and deploy without using WSAD and/or console ? I'm asking myself.
    - For now, it works but on the same machine : if i build on one box and the deploy on the other from my dev box : the deploy task (wsInstallApp) is not working saying that it cannot find a file : com.ibm.websphere.management.exception.AdminException: ADMA0043E: C:\app26673.ear (which is not my original ear) not found. Of course, don't expect to find any infos on this...
    1.3. Lauch the build process (WEBSPHERE SPECIFIC):
    Run ant task (wsejbdeploy) with your ear. On my 512MB machine it takes 450 (7-8 mn) seconds to complete. It makes me a websphereonly.ear, about 1.5 size of my original one. At this point I do not know how people that have dozens of ejb do, but i like to know if there is any. For me, it gives me the time to write this article !!!!
    Other server : one file copy is enough (1 sec to copy, 10 secs to deploy)
    1.4 Stop your application previous version (20s) (WEBSPHERE SPECIFIC)
    1.5 Desinstall it (20s) (WEBSPHERE SPECIFIC)
    1.6 Uncompress the generated ear, get the generated table.ddl that lies within and manage to run it against your database (destroy schema - drop tables, create a new one and so on..). About 3-4 minutes. (WEBSPHERE SPECIFIC)
    Any way of deploying database schema automatically, at least on cloudscape, which is shipped along ? didn't found it (this is the default on other appserver).
    Any way to build your ear independant of database back end ? didn't found it. For now I must ejbdeploy again (and again, and again...) - see step 1.3.
    What if my DB isn't in the IBM list ? Didn't found it.
    1.7 Install It (30 s) (WEBSPHERE SPECIFIC)
    1.8 Start It (20 s) (WEBSPHERE SPECIFIC)
    1.9 Stop and start your server (3-5 mns). This is because (i suppose) you don't know when you have to do it and when not. The console is saying : maybe you should restart your server : may be or may be not. To be sure, I restart.
    This is getting worse and worse if you plan to debug or turn on some logs or profiling. (WEBSPHERE SPECIFIC)

    TOTAL TIME : 15-20 MINUTES.

    2. Run It
    2.0 Buy a BIG BIG Box, even if you're the only client on your app, debugging. (WEBSPHERE SPECIFIC)
    2.1 It does not work. The SAME application is working UNMODIFIED on JBoss and weblogic, but not on websphere. For example have you tried to run local ejb calls in websphere ? If you want do it, you have to discover that there is a local namespace and use IN YOUR CODE. Since this, i have in my code (write once run anywhere BUT on websphere):
    (WEBSPHERE SPECIFIC)
    static final String WASLOCALJNDISTRING = "local:ejb/";
    public static String getSequenceLocalHome() {
    if (SystemProperties.isWebsphere())
       return WASLOCALJNDISTRING + SEQUENCE_HOME;
    return SEQUENCE_LOCAL_HOME;
    }

    Try to find where it is clealrly stated in documentation... You will have fun. And one bad try is expensive : see points 1.x !!!

    2.2 It does not work and there is mysterious errors happenings :
    - subtransactions not supported ??!?

    or some mysterious exceptions like:

    com.ibm.ejs.container.CreateFailureException: javax.ejb.TransactionRolledbackLocalException: javax.transaction.TransactionRolledbackException: com.ibm.websphere.csi.CSITransactionRolledbackException: null; nested exception is:
    javax.ejb.TransactionRolledbackLocalException: javax.transaction.TransactionRolledbackException: com.ibm.websphere.csi.CSITransactionRolledbackException: null; nested exception is:
    com.ibm.ejs.container.HomeDisabledException: OpenServer#sequencegenerator-ejb.jar#Sequence
    com.ibm.ejs.container.HomeDisabledException: OpenServer#sequencegenerator-ejb.jar#Sequence
    at com.icominfo.openserver.accesscontrol.provider.ejb.entity.EJSLocalCMPPrincipalHome_300c336f.create(Unknown Source)
    at test.com.icominfo.PropertyTestEJB.setUp(PropertyTestEJB.java:55)

    helpful, no ?

    Generally speaking, when it does not work you're in big big trouble because you WILL not find the cause of your error : neither in docs, neither on ibm website, neither on news (newsgroups on ibm app server are incredibly poor vs BEA ones - for example) --> (WEBSPHERE SPECIFIC)

    3. Try to debug
    If you put trace on, you should carrefully select whch one because you will have no longer a usable machine (it slows down so much that it is unusable).

    4. Call IBM support
    After five detailed mails to evaluation support, I do not have any response but "we're looking your problem". There is definitively NO evalutaion support. You will have to pay to see. For me, I would not ask my company to do it after this demonstration.

    5. Use WSAD... and stay on it. (WEBSPHERE SPECIFIC)
    It seems that a big point on IBM strategy is hiding proprietary details behind WSAD such making you completely dependant on it. IBM technical papers are WSAD screen shots compil. My understanding is that: if you're using WSAD things get a little less complicated but you're catch, exactly what IBM is wanting to keep his market share !!!

    So, at this point, even an IBM Global Services engineer should have switched to bea or jboss or whatever you like: I think it could not be worst.
    For now, I'm about to definitiveley remove this from my Hard Disk, despite the "market share" that IBM is claiming, unless some WAS fluent guy is pointing me where I'm completeley missing something !!! Have fun,

    Richard

    PS: if you would like any details to make the discussion go forward, don't hesitate.
  24. BEA Claims Big Wins Against IBM[ Go to top ]

    I think BEA & Websphere are both good app server. But the release of Websphere 4.03 has taken a good lead over WL 7.0. I have deployed application in WL 7.0 in production which is really very unstable and flacky on high load, The WAS 4.03 runs fine in high load and seems to be 30% or more faster than WLS 7.0 . I hate to say this but WLS 7.0 is still not ready for production. It has too many many issue on the console, node manager, clustering and run time issues I would recommended to stay away from WLS 7.0. I hope BEA will resolve the issues soon on the other hand Websphere console and ease of creating nodes and clustering beats far than WLS 7.0. The load balancing, scalability and hot deployment of WAS 4.03 clearly makes it a winner. I have been working on java for more 6 years and have tried both appserver. Websphere 5.0 will be another clear winner for IBM from what I have seen so far. I am also trying out the Oracle 9iAS and JDEV 9i and looks good but not mature yet. I am really glad that we have options to choose app server in the market
  25. BEA Claims Big Wins Against IBM[ Go to top ]

    hmm, just wondering. How does jboss stack up against those big guys ? Have you had any experience or heard anything about this *free* app server ? Thanks for the feedback.
  26. BEA Claims Big Wins Against IBM[ Go to top ]

    Yin: "How does jboss stack up against those big guys?"

    IMHO - JBoss is the easiest EJB platform for development and the most cost-effective (duh!) for bundling.

    My guess is you'll see most ISV apps developed for JBoss (low cost turn-key, default server), WebLogic and WebSphere (perceived commercial market leaders).

    Most large companies do internal "enterprise app" development solely on WebLogic, WebSphere, or another proprietary (in the objective sense) application server.

    Regarding performance, it really depends what you are doing. JBoss is easily "good enough" for the majority of applications, and in a few situations it meets/beats the "market leaders". However, in general, JBoss is not selected for performance reasons, but rather (a) ease of use, (b) availability of source, (c) ability to customize and (most obviously) (d) cost. In that sense, you could compare it to Apache Jakarta Tomcat, which you rarely select for speed reasons, but the performance is usually good enough, it usually is very quick to support the latest standards, it is ubiquitous in its development use and deployments, and the price is right.

    Peace,

    Cameron Purdy
    Tangosol, Inc.

  27. IBM Disputes BEA Application Server Claims
    August 19, 2002

    http://www.eweek.com/article2/0,3959,476252,00.asp

  28. BEA Claims Big Wins Against IBM[ Go to top ]

    I have a question for the BEA employees..

    How are the boys from Redmond (x-microserfs), whom you have employed doing? I see they have been busy beavers propelling BEA to non-standards J2EE.

     
  29. IBM Disputes BEA Application Server Claims[ Go to top ]

    http://www.eweek.com/article2/0,3959,476252,00.asp