Giga: IBM nearly matching BEA in marketshare


News: Giga: IBM nearly matching BEA in marketshare

  1. A recent press release brings new data from industry research firm Giga Information Group. According to Giga,
    BEA's market share increases from 32% to 35% but IBM doubles from 16% to 30% bringing it neck and neck with BEA. These numbers are based on survey of 52 customers.

    Press Release:

    BEA shares plunge on analyst's concerns

    NEW YORK, May 4 (Reuters) - Shares of software maker BEA Systems Inc. (NasdaqNM:BEAS - news) on Friday slumped more than 10 percent after Deutsche Banc Alex Brown analyst Jim Moore issued a cautious outlook on the company past the first quarter.

    At midday, shares of BEA were off $4.11 to $35.75. The shares were among the most actively traded on the Nasdaq and among its top three net losers.

    Moore said he was concerned that several factors may hurt the company's high share price valuation.

    ''Valuation is still our major concern, and we urge investors to be selective in timing,'' Moore wrote in a research note.

    Moore said he was comfortable that BEA, which makes software used as a foundation for program building, would meet or slightly beat his estimate for the company's income of $253 million, or 7 cents a share, when the company announces its fiscal first-quarter results on May 17.

    However, he said he has learned the company may be cutting sales quotas and that only about half of its force met their sales quotas for the first quarter.

    He also said he is concerned that BEA's components business, which makes up 15 percent of its revenues, was down 40 percent in the April quarter. Additionally, Moore wrote that any slowdown in new applications development projects, especially in the high-tech, financial or telecommunications sectors, would cloud visibility into future quarters.

    Finally, Giga Information Group said IBM is gaining ground, with BEA's market share rising to 35 percent from 32 percent, while International Business Machines Corp.'s (NYSE:IBM - news) share nearly doubled to 30 percent from 16 percent, according to the research and consultant firm's survey of 52 customers. BEA and IBM are the leaders in what Moore estimates to be a $1.64 billion market.

    IBM's rise may signal the competitive landscape is heating up. Still, BEA is scheduled to unveil some new products during the summer, which could open new opportunities in the growing portal market, Moore said.

    Threaded Messages (32)

  2. Wow,

    They surveyd 52 customers.
    What an impressive figure.

    The same bunch of morons who don't know anything about what they're talking about when they make technology predictions and recommendation and when it comes to statistics they survey 52 percent of thier customer.

    This statistical sample is indeed relevant for the Giga customers category.

    But they failed to mention what was the statistical sample when they took the previous survey.

    Another joke from Giga, unfortunately these guys have some influence in media circles as well as CXO's circle.

    But I think here we are a professionals community so ... them.
  3. Hello, I'm the author of the Giga report cited (and misquoted) in this financial analyst's report. The market share data is NOT based on a survey of 52 customers!!!!

    Market share data is ALWAYS based on revenue. Period. That's how everybody does it. The trick is finding out the numbers, which we have sources to verify from multiple directions. We size this as a 1.64 billion dollar market (in 2000), so you can work out the numbers to get to 30% and 35% share for IBM and BEA, respectively. These numbers include both product and service revenue. And nobody else even comes close. The closest is iPlanet with a 9% share.

    Later in the same document, I presented an analysis of 52 Giga clients (mostly very large companies) who have bought application servers, as further evidence and to provide a cross-check. Of this group who bought BEA or IBM (and very few bought anything else!), 54% bought BEA, 36% bought IBM, and 10% bought both. I've seen survey results in other forums like the Cutter Consortium that saw an even higher result for IBM. Clearly BEA is the leader, and also has the best product (as covered in our report, too). But IBM has made significant headway in 2000 revenue over 1999, as reflected in the market share gain.

    Note that in the first three quarters of 1999, IBM's product wasn't really usable, it only reached saleable condition in Q4-1999. Whereas in 2000 they had saleable product all year. This is the main reason for the significant change in share. Most IBM customers would prefer to buy WebSphere if they can make it work, and that generates a lot of sales for IBM, which also has advantages in some areas including service/support, development tools, and related application integration products. IBM has also made significant product improvements, and with WebSphere 4.0 is likely to be functionally much closer to BEA. And after all, as you can see from other threads about BEA in this same site, BEA isn't perfect, either!

    Also note that we have taken several surveys of attendees of Giga conferences which have shown similar levels of uptake, BEA #1, IBM #2 but fairly close behind, everyone else way back in the weeds. Some of those other products are very worthwhile, don't get me wrong, but they don't have significant market share. For some buyers, that's an important part of the decision framework. But we also recommend a careful consideration of requirements, and mapping into the functional characteristics of the products. We also provide detailed evaluation criteria, and our own assessments. But this site is also a very useful source of information, very consistent with information I get from other sources (such as our many clients who tell me about what they're doing with these products).

    I hope this clarifies this unfortunate misrepresentation of our research.
  4. Mike, thank you for the clarification! It puts to rest the rumor mill.

    (And I'll second that comment about WebSphere being unusable during most of 1999.. :D)

    Billy: regarding entity beans: generally agreed, I don't use them, though I still have hope for them (or JDO) as a means to get O/R mapping out to the masses.

    Regarding VisualAge (I'll keep this brief)

    I've used VAJ Pro since the 1.0 , and VA Enterprise 3.5 for the last year and generally I've been pleased except for the lagged JDK versions and the recent version's level of bugs (makes me want to move back to Canada to work at IBM Toronto Labs & get that product back in shape... <grin>).

    In short, NO Java IDE comes even close to offering the searching, source control, and refactoring (with jFactor) capabilities of VisualAge. It has its quirks, it has its moles, and most of all, it has a big learning curve. But every time I switch away, (and I've tried pretty much every Java IDE out there for months at a time) -- I have to go back to VisualAge if I want to perform a dependencies analysis, or if I have a major interface refactoring, or if I don't want to have to write all of my get/set methods.

    While at a client site I always recommend it with the caveat that one needs training / mentoring to use it effectively. At my last engagement this went quite well for some projects, and not for others (those with external source management).

    IBM has a great tool on their hands, I hope they continue to exploit it (and don't take too long to release 4.0, apparently it will be a re-write)...
  5. Yup,
    I completely agree with the above comments. I have used almost all IDE's in the market & see nothing close to VAJ.
    I love to work with JBuilder & also i am astonished to see its performance inspite of "All Java" inside it. But when it comes to serious programming VAJ is best.
    I rank VAJ becoz of the following reasons,
    a) Version control
    b) Code Format ( It helps to me to easily review any of collegues code )
    c) Shows only the method as a whole so that better abstraction of the code
    d) Build-In Websphere test environment.
    not to mention "auto complete" etc etc
    VAJ is the best.

  6. I could bring my personal experience with all the IDE's, and VisualAge first for Smalltalk ( still the unbeatable) and then for Java are a lighting year ahead from others.

    Other IDEs are simple editors in comparison.
    I do not want to talk about all the features that Enterprise edition has, nor the powerful source code analizer.
    Do not even talk about the persistence builder feature
    ..a complete visual O/R mapping framework for simple data object or EJBs. The visual builder is great. You can rebuild the GUI given a non VA Java source code
    The debugger is just amazing. WAS runs inside VA4J,
    deployment of EJBs piece of cake with a click!
    Now I use JBuilder because it is the tool my company uses,
    and I can tell you I miss Va4J

    If I want a personal IDEs for home, I would use Jcreator 2, free, code completition, debugging interface etc.. ( I would not spend thousands of $ for others.
    If I want a enterprise team IDE, I have no doubts! Va4J
    Unfortunately, nothing is pefect. Va4J works well with its built in repository (by the way, its atomicity is at the method level!! ) but does not with external CVS properly
    You cannot plug in your JDK. You have to use the IBM built in JVM, which is the worst thing I could say about it.
  7. I guess that using revenue in deciding
    market share is logical in a way since that
    is what the big players (BEA, IBM) is living
     But will this not lead to some rather strange
    results when the Open Source players are taken
    into account? In an extreme case 90 percent
    of all firms could be using Apache/Tomcat/JBoss
    or some other free configuration but that
    would still give an almost zero market share
    revenue vise. (You could get a bit above zero
    for those buying open source support)
  8. If we're talking market share, why is <b>services</b> revenue counted? It seems to make sense to me that if your only metric is revenue, it should solely be <b>product license</b> revenue alone. Service revenue is a completely separate matter.
  9. The problem with using revenue as your yard-stick for measuring market share is game this lets companies like IBM, which crosses many IT product markets and bundles products almost as much as Microsoft. Every iSeries server (AS/400) comes with WAS, and be sure that at least a portion of that revenue is counted as WAS revenue. Every VAJ comes with a stripped down WAS for testing, so be sure that that is counted for WAS revenue as well.
    A better tally, if possible, would be production system installed base. Probably not doable, but as long as you use revenue to measure market share, IBM and companies like it (Microsoft for one) will play these games.
  10. I am not surprised at all: last year IBM announced they would spend a billion dollars marketing WebSphere. Working for an enterprise which choose WebSphere I think could a lot of big enterprises will choose for IBM as they always did...... (a well known saying: no manager ever got
    fired for choosing IBM). Because of this any non IBM implementation of J2EE (and also .Net) will have a hard time in competing with WebSphere in the big enterpise
    arena (I don't it is a technical issue but a political).
  11. IBM will get large marketshare (assuming they don't put out a totally faulty product) because of the large IBM shops. They have all the pieces to the technical infrastructure eg. MQ , connectors, etc. Why go to a competitor ie BEA when you can get it from a single vendor ? Current surveys indicate that EJB is taking time to be accepted and so this works out in IBM's favor. With many companies preferring single vendor and not being in any rush (now that the .com 'revolution' has stopped to a standstill) I expect that IBM will clearly be at least #2.

    52 companies ? Which one ? Big ones ? Ones that have been around a long time ? Guess what vendor's hardware and software they probably run (at very least their legacy systems)? Integration of new and old applications is going to be very important to them..

  12. Are they kidding?

    52 is not a sample statistic, it is a population parameter. It is *too low* to be of any statistical significance. And to think people base their decisions on such poor quality numbers like this. Geez, we have 52 developers using various AppServers for various projects here in my own place of work alone (consultancy). Does this really mean anything worthwhile?? NO!

    Such a shame...

  13. Either way it's all scary if IBM is increasing market share. Getting a job in Websphere having used a more compliant container like Weblogic is very difficult because it also requires extensive experience with Visual Age. IBM's Websphere product has significantly fragemented the J2EE marketplace. Unfortunately IBM has a lot of marketing clout in Europe also and many roles there are also Websphere only.
  14. Well,
    Every survey needs to be taken with a piece of salt. We all know the stories behind the last one giving BEA 54%. Once, people realized what happened there then these results get no more credence than benchmarks.

    But, to be honest, as regards the comments regarding IBM fragmenting the market, trust me (at the risk of being stoned!), J2EE compatiblity isn't as important as people think. The software stacks being built on top of J2EE base products by BEA, IBM, Iona etc will kill any pretentions of portability anyway. It's going to take Sun some time, if ever, to add standards for all this in to J*EE.

    Very few of the big vendors have said they will cross-sell their stack on other J2EE servers. It's only the smaller vendors that have said they'll do this.

    I think a J2EE servers role in an enterprise deal is less significant than most developers realize and it's the added stack that is the differentiator. This basically locks you in anyway.

    As regards compliance, WAS 4.0 when it ships will silence most critics (albeit very late in the game). You'll finally get J2EE compliance, 1.3 and the VAJ requirement should be gone if you don't want to use it. I'm currently using NetBeans, WAS 3.5.3, CVS and ANT for a full development environment. I'm running it all using the latest IBM JDK 1.3, debugging with JPDA (NetBeans). The one thing I still can't do is CMP cause the tooling is built in to VAJ (this changes in 4.0). This goes to show that whilst it's not documented very well, you don't need VAJ for WAS development although it certainly makes things easier.
  15. The IBM press machine is constantly working in high gear. Spending millions on TV ads, print ads, market surveys and press releases. The feedback I've received from developers is that unless you're an IBM shop forget about WebSphere. But there are many IBM only shops and they're now developing applications.

    How many of the 52 in the survey are IBM shops?
  16. If this is not a joke, then there can be nothing called a Joke...Giga releasing reports based on a statistic of 52!! i guess the they mistook the appserver user to be a wild african elephant(a rare species i guess)...I hope that they realise that there are more than 10000 times that many...IBM websphere is still not EJB1.1 compliant....i am sure no developer community would love to embrace something as passe as this....hope people bring out better reports guys...
  17. I think that J2EE compatibility is quite significant at least for the next 2 years. These "software stacks" don't exist yet in volume, and of those that do are often of dubious quality or merit.

    IBM's marketing is certainly working; let's hope the technology follows suit. I love VisualAge, I'd almost use nothing but VisualAge except in peculiar circumstances, so the EJB environment in there + websphere would be a nice fit for me.

    I personally would love to see more value-added software, but I've found the J2EE industry in general to be very afraid to innovate / venture outside of the spec. The dearth of good CMP implementations is an example of this.
  18. Hi Stu,
    I agree with you on VAJ although we're probably the only two on this portal that would say that. I use VAJ when I can. The only trouble with it is that it requires a fair bit of training before you realize it's cool and a lot of developers never do this and hence hate it.

    I dunno about innovation outside the spec. WebLogic, for example, is full of non-standard APIs, look at the weblogic.* packages. This is all documented and supported but is outside the spec. The CMP problem is I think more to do with the difficulty in doing a really good one. It's a tough problem and I'd very rarely even think about using entity beans except for prototyping. It's been my experience that you can use them for prototypes/initial versions as a development time saver but when the times comes to start tuning then they are the first things to be converted to SQLJ/JDBC because of performance issues.

    But, I meant innovation from the point of view of layering middleware/applications on top of the J2EE server and I think this is starting to happen now. Silverstream, BEA, Iona and IBM. I agree about the merit/quality aspect right now but it's going to get better pretty fast I think.

    You'll have to excuse me for my comment on the significance of J2EE. It was only meant from the enterprise customer point of view. I'll clarify. The standard is definitely a good thing. JCA is one of the more useful things in it right now, for example. But, enterprise customers need a lot more than this to build enterprise sorts of applications. BPM, B2B, single sign on security solutions, messaging, message/service brokers, interoperability with legacy systems (probably messaging but also JCA).

    But, as groups start to use more of the extras then whether the J2EE server is Vx.y versus Vz.w becomes less important. You have to use the one that the particular stack mandates not the one you want to use. This point is familiar to anyone using a stack of any kind right now, the matrix of versions of what combination does work versus what combination is not support is usually a short one. It's rare that you use the latest and greatest version of anything except the topmost component in the stack.

  19. I think that J2EE compatibility is quite significant at least for the next 2 years.

    Yes, I bet it is.
    The clumsiness of the EJB specification will weigh heavily like a ball and chain on the capability of app server vendor of making a quality, innovative product.

    The monotony of heaving every potential innovation in App Server area waiting for the next specification release and eventually be dropped because of compatibility issues will also have a terrible impact on the quality you mentioned.

    Common, as long as the main product follows the general J2EE line because of the necessity of getting J2EE certified, the innovation you'll see in com.bea.* or com.genstone.* or whatever, will always be minor.

    The first App Server vendor that will have the courage to choose NOT to be J2EE certified will be my favourite and I'll recommend it for everyone who asks me.

  20. I dunno about innovation outside the spec. WebLogic, for > example, is full of non-standard APIs, look at the

    > weblogic.* packages.

    Er... do you prefer "*"? Because if it's just a matter of naming, it's easy to change.

    I can assure you that weblogic.ejb20.* is 100% EJB 1.1 compliant (and as close to the EJB 2.0 specification as it gets.

  21. Cedric,
    If only the package was weblogic.ejb20 then fine but that's not the case and you know it. IBM has very few documented classes and I think it's a shame!

    BEA have documented quite a few useful proprietary classes and the product is better for it. IBM has most of these internally but unfortunately chose not to document them and I have been screaming at them to change this policy.

    J2EE servers need proprietary extensions to make them useful. J2EE isn't perfect or mature enough for complex applications. BEA recognised this, IBM did internally but didn't document it. One up for BEA.

    Do BEA track the standard with the implementation better? absolutely.
  22. If only the package was weblogic.ejb20

    > then fine but that's not the case
    > and you know it.

    Billy, I'm sorry but I absolutely don't see what you are referring to. All our EJB 1.1 and 2.0 container lives in the weblogic.ejb20 package, I don't understand what disturbs you about that?

  23. Cedric,
    Don't take it the wrong way. There are WL startup classes, WL trace classes, all the Toplink CMP stuff for making it useful. That's what I meant and I don't think there is anything wrong with that in any way. These 'proprietary' but very useful extensions enhance the product and don't detract from it.

  24. - Being a developer it is difficult for me to see how IBM could be having 30 percent of the market share. This is because IBM recently released its so called ejb 1.1 compliant Websphere version 3.5 in its production stage(3 months). The version before that was not capable of running in java 2 environment. The appserver has lots of basic bugs like connection pooling, webserver appserver integration....
    - It is very difficult to integrate IBM Websphere with any other application.
    - It was the one of the last server vendor to make websphere ejb 1.1 compliant. I dont say this is the case for all IBM software. So you might want to give it a thought before you go in for that.
    - IBM which is supposed to work best with DB2 itself is buggy in its 7.1 release.
    I can add many more points but i rest my case here.
  25. Has anyone thought that the 52 companies might represent companies who buy the *most* web app. licenses relatively speaking. If we had a survey of a 100 companies and those companies only bought < 5 licenses, they really don't represent large and medium purchases.
  26. Well, WAS 3.5 isn't 1.1 compliant. 4.0 will be. 3.5 is 1.0 plus RMI/IIOP and CMP. 3.5 has been out for over a year now (Oct 2000). V3.0x was indeed runnning on top of 1.1.8. So, your facts are a little topsy.

    We've deployed several applications that used connection pooling and talked to both Apache and NES as the web server. It isn't perfect but you're a little to negative I think. Most of the problems developers had were related to VAJ and the development tooling and boy, there were problems :-).

    Once it got to production it worked and IBM were reasonably responsive in fixing S1 problems. There are a lot of production applications running WAS at this point including some very well known ones including everyones favorite, Victorias Secret.

    It's difficult to integrate an app server from any vendor in to an existing environment. What sort of environment are you imagining? A true legacy environment or a Java environment? IBM made available its CCF toolkit since V3.0x for legacy integration and CCF was basically the one of the main precursors of JCA which incidently IBM now has a beta shipping with quite a few JCA adapters.

    IBM has learned from its mistakes as regards J2EE compliance and I think we'll see improvements now in this regard.

    V1.x of any product has its share of bugs. DB2 V7.1 has, of course, its share of them, as does WebLogic 6.0, Oracle 8i V1 (8.1.x) when it came and so on.

    I am usually the first and sometimes one of the loudest to shout about the problems with WAS/VAJ (ask anybody at IBM) but I believe in constructive criticism rather than just saying everything is bad.
  27. You hit the point IBM Websphere isnt 1.1 compliant. It takes approximately one year to release any software's production release by IBM. Considering that it will be next year when Websphere 4.0 production release is implemented. Till then the world will be far beyond the EJB1.1 spec.
    Which is the only development environment where you can create EJB's for Websphere.
    Surprise Surprise it is IBM Visual Age.
    With a set of 500 classes in its project it crashes once in two months taking everything with it.
    Because it loads everything into memory even the deltas of any class ever loaded. And where do all these class and java files reside. In one vaj.dat file. How large can this file get. Approximately 200mb or larger with more than 500 fragments in a 15 Gigabyte hard drive.
    Does Websphere have a deployment descriptor. No.
    What webserver does Websphere use.
    Consultants say that it is a modification of Apache web server.
    About victorias secret. If victorias secret is running on websphere it is ok because most of the pages dont even hit the appserver and apache takes care of application.
    But when a page does hit the appserver hopefully through the servlet engine which might be a couple of transactions in the entire website. IF you tell me that EBay is using Websphere you know i would be impressed. I am glad that you know that Websphere 3.0 fix pack 6 runs on ibm jdk1.1.8 which does not run on any profiling tool what so ever on earth.
    Guess which is the latest fixpack for the production release of websphere3.5 in its release of 2 months and some days. It is already fixpack 3. Does it run with ibm profiling tool. I hope you know which one. It takes 512 megabytes memory to profile approximately 50 classes with 10 ejbs. That too session ejbs.
    Parts of Websphere and Visual Age both are written in small talk. IBM is trying to rewrite the entire appserver for its 4.0 release. I wish them luck with that. Please let me know your thoughts.
  28. In past, you'll dead on. IBM took way too long to implemenet EJB 1.1, I'm hoping that with 4.0 that will improve and they will track the standard more closely. But, that hope, we'll wait and see. As for 1 year, they are no different than anybody else. Is WL 6.0 production ready? No. How long did WL 5.1 take to shake out? A while. All products need a shakeout time (a shame) but that's life. At least BEA had the excuse for the bugs because they tracked the spec very closely. IBM didn't even have that excuse. I make no apologies for IBM. I too have suffered as they say.

    You only need VAJ for CMP development. You can use any IDE you want to build beans for WAS right now. I use NetBeans/JBuilder and ANT and that gives me a full dev/deploy environment from the command line. Losing CMP isn't a big deal for me because I don't think they are a liability at this time.

    I don't have your problems with VAJ (I have other problems of course) but overall I like it and my repository is 700MB.

    It doesn't load every fragment in RAM, they are all stored in the repository and I think it's a good thing and one of VAJ's strong points, the auto version control. I just wish IBM bundled tools like VA Assist with it to add the nice things to make envy a lot easier to use.

    WAS has XML descriptors and has had them for some time. Look at the jetace tool with the -f option. You'll find what you're looking for. They aren't 1.1 compliant but they are there and they're functional.

    WAS uses almost any web server you want. The only issue is that they may not always support the latest service pack of your web server but it's the same with everyone.

    I would hope that any web site works as you describe. The fastest sites won't go through a dynamic engine for most page requests, they'll try to use a static page that can be served by a web server. This lets it scale well (add cache boxes/web servers is cheaper than adding boxes with app servers). If your sites don't work like this then I'd say you should look closely at what you're doing. Don't take this as criticism but as a rule, we should aim to deliver static content when possible for performance reasons. You can then leverage caching systems like Akamai much more readily to scale your site.

    WAS 3.0x supports JOptimize it for as long as I can remember so there is your profiling tool.

    WAS 3.5 is indeed on FP3, WebLogic 5.1 is on SP8. Whats your point?

    It does take a lot of RAM, too much, I hope they sort that out.

    I'm not going to get in to a 'me defending WAS'. That's not my intent. IBM are 'big and ugly' enough to stick up for them selves. But, I use WAS every day and have used it for around 3 years now. Yep, it absolutely has problems and I'm the first to point them out but you can built stuff with it and it does work.
  29. Hey Bill,how many people on earth would like to work with a 700 MB file which has more than 50 fragments on your hard drive, where if one bite goes wrong all your work is gone and you may have to spend an entire day installing and configuring software not having a single class of your code. I hope that you keep your code some place besides IBM development environment, mercy one byte goes wrong in in the IVJ.dat on your file system. If you do like it you are exceptional.

    JOptimize goes up to profiling the servlet engine but when it gets to the appserver(EJBs) it simply kills. it. If you think thats profiling I am again sure that you dont develop full time. Besides one of the profiling tools I use for IBM stuff reveals memory leaks in IBM code, scary isnt it, not one or two classes in ibm but tens of them.

    I am not criticizing IBM but these are facts. It just needs work. I am not a fan of J2EE standards either. Mainly because the whole concept of Entity Bean somehow doesnt appeal to me. But I love session beans. But even without j2ee standards I would be happy if something just works you know. But here we have neither you know. Doesnt work and no compliance.

    I recently seen an IBM advertisement saying that when you write code for IBM Websphere you write 80% lesser infrastructure code. Then why is it that when you write classes in JBuilder do you have to write Ant scripts to deploy EJBs and set up the environment. And when you write code in JBuilder you have to configure jbuilder to use IBM Jdk's and Jbuilder doesnt work so well with IBM JDK you know. Debugging becomes a nightmare. I hope u understand why we have to use IBM jdk if we have to work with Jbuilder and use ant script to deploy code into websphere.

    I understand that you like ibm and it works for you. Good for you but not good enough for me. I prefer other vendor servers.

    Just to let you know that IBM Websphere is one of the most expensive (Enterprise edition) with some of the least functionality that other app servers provide. I would be happy to know if IBM appserver provides anything besides CMP, connection pooling, object pooling and name service.

    And hey, Im just a critic, Im not in favour of any appserver. I am just revealing facts for people to know about this App Server by IBM.
  30. 700MB or 1Mb, what's the difference. It's a question of setting it up right. We've have teams using envy for several years now, multi-site replication and never lost anything. Nothing was lost. I think you're just scare mongering here. Yes, it's different but that doesn't mean its worse. For server development I'd take VAJ anytime but for client development, especially GUI clients then it's got big limitations.

    Put the Team Server on a good quality box that gets regularly backed up and then forget about it. Just get the admins to compact it occasionally. Do people stop using a database because it's too risky because there is 1gb of data now and one byte wrong screws the database? I don't think so.

    We've a different experience to you, we have applications in production using session beans, BMP even some CMP (not for me though) and it works. We've had problems with clustering but they seem to be resolved now. What you seem to be saying is that you couldn't get it to work. Thats different than saying 'It doesnt work'. It could be a lot easier to use at the same time and I have many bumps on my head from banging it of the nearest wall occasionally!

    The biggest issues with WAS are the tooling (admin console, xmlconfig), the memory foot print being high, integration with VAJ at the expense of other IDEs, the need to use IBM's JDK, as well as not enough proprietary features that go beyond the spec, i.e. add value on top of the spec.

    As for JBuilder, with consultants or outsourced development etc on development teams, sometimes it's easier to let them use the IDE they want and insist on a standard non IDE specific build environment (version control + build). WAS works best with VAJ, no question. That's my first recommendation, if they won't do it then it's a question of what can you live with. JBuilder didn't like IBM's JDK at all in the past, recently, we run it on IBMs 1.3 JDK on NT for development/debugging and it works well enough for clients although server debugging is still flaky.

    As for pricing, I don't agree. WAS seems to be around 10K/cpu right now. WL is 13-14K and once you start adding in other features in almost anyones (Iona, IBM, BEA, Sybase) stack you're looking at 20-40K/CPU easy. Now, add on support and development costs of implementing a couple of projects and trust me, the product licenses aren't such a big deal anymore in any case.

    WebSphere Enterprise has some features (the persistency layer) that are/were leading edge but the market for such features was tiny hence it's marked lack of success.

    I don't 'like' IBM, it's just what I use day to day and hence have most experience with. How-ever, you never know what the next client will ask you to use in the consulting game....
  31. Hey Bill why should I use team developmet server to store my code and buy another software from IBM, another repository by paying money to ibm. Its too bad you dont see a difference in a 700 mb file and a 1 mb file.

    You say that JBuilder runs with ibmjdk 1.3 but websphere still runs with 1.2.2 some fix pack. If you are doing development in jbuilder on ibmjdk1.3 then you are not writing code for enterprise edition for the websphere appserver. Please do not confuse the readers here on which jdk the websphere environment runs here. And with jdk1.2.2 code it doesnt work with JBuilder. So in effect what I can see is JBuilder still cannot be used for Enterprise Java development for Websphere because IBMJDK1.2.2 doesnt run on JBuilder so well or vice versa.

    With the fixpack you hit the point again. IBM has only 3 fix packs while other companies constantly keep updating their servers to do better. Why?

    The servlet engine dies under load and also messes up sessions in the memory.

    IBM has a load balancer for distributing requests of the webserver. But that is a little buggy and takes forever to configure.

    The pricing of IBM Appserver isnt a big deal. But it is very limited in its functionality for what you get with the same pricing from other servers.

    I am happy that clustering works for you. But after seeing the installation and configuring document for doing clustering and the missing information in the documentation we had to give up on clustering for Websphere. I guess you must be in really close contact with IBM as about a couple of months ago the clustering was not possible.
  32. Hey Sid,
    If you're using VAJ then you should use Envy. Trying to make VAJ work with another external SCCS system is just going to cause grief and 'life is too short!' as I always say.

    If you use envy then you need to use it the 'right' way. This may sound nuts but the way it is used is very similar to the way you use high end version control products and believe it or not it makes sense when used like that. How-ever, if you try to use it like CVS etc then you'll run in to problems, thats not the way it works and you'll end up in knots. It's a powerful tool but without the right training and know how then it's a frustrating partner.

    If you don't want to use VAJ then go with the command line build environment and forgo CMP. If you do the non VAJ route then if you need to debug then switch to IBM 1.3 for the debugging session although only 1.2.2 is supported by IBM and I'd only recommend running WAS 3.5.3 on the 1.2.2 JDK in production.

    Officially, like you say only 1.2.2 is supported with WAS. How-ever, it runs very well (although unsupported) on IBMs 1.3. You can also get SR11 of the IBM JDK 1.2.2 (ask your WebSphere rep for it, you can't download it directly) and it apparently works with JPDA also but I haven't tested this.

    IBM releases a fix pack every quarter for WAS. 3.5 came out in October, we're now on FP3. They release efixes in the interim for any serious problems. Efixes are normally rolled up in the subsequent fix packs and regression tested. This should explain your number of fixpacks question.

    WAS 3.5.2 had a few problems with clustering on the HTTP side. 3.5.3 fixes these issues. 3.5.3 only came out within the past couple of months so that tallies with your comment. Yes, the docs were poor but with the redbooks etc arriving now a lot of the documentation is there but it is admittedly hard to navigate around.

    Anyway, this isn't intended to be a review of WAS although it's beginning to sound like one. My only reason for spending time on these posts is merely to provide a balance for your comments from someone who has used WAS almost constantly for a couple of years now.

  33. GIGA's report is at lease irresponsible, and at worst intentionally misleading, even fictional. This is no suprise, and anything this completely vendor-influenced think-tank has to say should be scrutinized or ignored altogether.