In response to the recent report by the Gartner group about overspending on app. servers, companies are beginning to seriously consider scaling back their deployments. In particular, TheStreet.com is migrating their Sun 250/450 servers running ATG Dynamo and Vignette to Compaq/Dell Win2K machines running the low cost JRun J2EE app. server and Apache web server.
This industry reality check should be good news for companies like Macromedia and Unify, both of which have the industries cheapest J2EE servers, and for open source efforts such as JBoss.
Read E-businesses are buying more application server than they need
And its going to get worse moving forward. I fully expect to see corporates consolidating boxes AND J2EE licenses to reduce costs. I think companies that did large single shot CPU license deals will start leveraging those licenses by deploying multiple applications per box rather than an application per box as may have been the case before as well as deploying the applications on more cost effective platforms such as Intel/Win32/Linux. This goes to the heart of our previous data center article that we published a couple of months back.
I don't think it's so bad for the bigger guys as regards large corporate customers. If a company has decided that WebLogic or WebSphere is the corporate standard then given the problems of deploying something else in production (training, support etc) I don't see them jumping to JRun or the like. Products like WAS Single server edition and strangely enough WL 6.x are light enough already. The license cost differences between these and the competition when compared with the retraining cost isn't enough to justify the 'hassle' of running a different server. I expect them instead to fund farms of machines running the corporate standard. Teams what need an application deployed will instead leverage that farm rather than getting their own pair of boxes as they did before. They should allow companie s to save quite a bit of money as regards box costs, box support costs, and J2EE license costs without changing J2EE vendor.
Now, for smaller customers that aren't so bothered about this standard and traing costs then it's different of course. If nothing else then it may cause the bigger vendors to price down their low-end offering such as WebSphere single server. Given the well publicised difficulty that companies betting on large deal sizes are having right now, I fully expect the vendors to start concentrating on getting the right price points for these smaller sized deals. IBM are starting to position WebSphere for this although I think the price is too high still. WebLogic don't really have a value J2EE server that I know of (I'm happy to be corrected here folks!). Their express product is basically just a servlet engine as far as I remember and this is an ace in the hole for IBM as I see it now.
M$oft FUD... I've seen this so manny times ... Microslobt is just trying to stem the J2EE tide. Seems Billy boy cant change his childish behaviour.
I Say no more !
Where did that comment come from? Who said anything 'bad' about J2EE here or that Microsoft are behind this.
As far as I can see MS only get a mention at all because TheStreet switched to W2K from Solaris but whats that got to do with 'stem the J2EE tide'?
With logic like this, maybe we all say no more...
In addition to the impact on large and small end-user
customers, a trend towards a common J2EE infrastructure
would likely impact ISVs, such as application vendors.
The latter would be hard pressed to migrate their apps
to run on top of the leading J2EE application servers
and abandon their proprietary infrastructure. Beyond
that, it would be interesting to see what pricing models
would emerge for these products which used to be server
based but now lacking their own run-time, are positioned
as frameworks and/or tools running on J2EE infrastructure.
Cost is a huge factor these days. Considering cost, I have persuaded my client to use open source JBOSS app server for my current project. I am very happy with JBOSS so far.
One appserver that really scales and performs I would say even better than Weblogic is IONA's iPortal Application Server. I did some evaluation work for my company and I was happily surprised that this guy really performs and cost wise its dirt cheap as compared to Weblogic or Webshpere. The guys in Dublin have really done hell of a job. We were earlier using Weblogic, which we have now decided to replace with iPortal Application Server.
I work for a .com, and to save cash we returned our entire Sun architecture (almost) and replaced it with Intel boxes, and consolidated several software licences.
Since we've payed for Weblogic, we probably won't ditch them, but we wouldn't purchase new licences at this time.
Cutting waste is a reality for most tech companies, and not a sign of impending doom for J2EE.
It will be interesting to see if application servers can avoid be turned into a commodity. What can Weblogic/Websphere add that the average application developer needs that can't be provided by free or low-cost app servers? Without adding ungodly complexity and hardware requirements?
Management tools would be one differentiator for the high end market. If (and this probably isn't an if any more) big corporates start rationalizing their use of J2EE through box/license consolidation then the ease of which they can manage N applications running on a large box becomes a big factor in picking a J2EE product. Factors such as providing quality of service guarantees to these applications become important.
I think this is an area that the bigger boys may leverage in order to convince corporates as to why they should choose them.
Perhaps this is really bad, and companies are discovering that the J2EE Platform doesn't really deliver the low-cost, scalable applications they want.
On the other hand, Jrun, a great product for some time now, might actually deliver the sort of speed and performance J2EE so often misses out on.
We might even see some convergence between ColdFusion and Jrun, dramatically increasing the pool of talent in the J2EE community. [And the amount of projects we can work with].
The big unanswered question remains whether they're just diminishing their involvement with J2EE full-stop.
In some ways, you have to ask yourself why they didn't do this to start with.
The rule of thumb I always give out is: build a scalable product, then deploy it small. If it turns out that the hardware or the app server you picked originally is underpowered, scale up! That's why you build a scalable app.
A point in Gartner's report is that 60% of J2EE applications use JSP/Servlets only without EJB.
If that is the case, you have many choices of servlet containers.
However, most of the lower-priced servlet containers do not have many bells and whistles of high-end applications, like clustering, user-friendly admin tool, etc. Unfortunately, moet of the high-end app servers, while having these niceties, also come with EJB and are priced much higher because of that - not clear whether the high prices are attributed to support for EJB or other benefits beyond the spec.
What are needed are simple servlet/JSP containers with the some of the good features of Weblogic - many of which are lacking in JBoss and JRun. OF course, the prices has to be reasonable.
Does anyone ever track what these bozos say? A year ago Gartner group and others said that BEA and IBM where the only way to go because they where the market leaders. I tried to argue for less expensive solutions, since the company couldn’t afford BEA, but know how that goes …can’t argue with the “experts”
Well now it’s a year later and Gartner complete contradict their realer recommendations.
But now I have a great time using JBoss so what the hec, I’ll use the Gartner report to prove to management what a smart decision we made.
Rexip AppServer is just the right product that you are looking for: servlet containers with many bells and whistles of Weblogic but without the overhead of EJB. Please take a look at www.rexip.com and get a free download.
We will annouce the pricing in a few more days. I can just say at this point it will be comparable to other servlet containers (but with many features of high-end app servers - including clustering with in-memory replication with which you can piece together many PC's to achieve performance better than a Sun workstation, yet with lower cost and better reliability). Rexip AppServer has been used by banking, mutual fund and other financial factors. It is also popular among trading companies. We also have OEM relationships with indepedent software vendors who develop applications on top of the low-cost high-end application server.
Send me an e-mail if you want to know more.
niles at tccybersoft dot com
I wonder if those high-end appserver vendor will price down their products. That will just cause them too much because that means cut down in revenue and even stock price as investor will question where do they get new revenue streams...
mmmm this is an interesting situation..Atlast the "Cost Control Sword" has got into H/W & S/W arena !!!
This situation will make appserver Vendors to price their product competitively .. So who are the gainers & loosers ?
Who will gain the maximum ?
Linux & lightweight appserver vendors. Now these vendors will have some breathing space after a heavy crush by heavyweights like Weblogic & Websphere .
Who will be the looser ?
a) Obviously heavy weights !!.Till they come with a pricing model which charges the client depending on the services used,instead of per CPU etc.
b) Microsoft !!! Yes.. now if these lightweight vendors prove & deliver cost effective solutions to the market ,Microsoft cannot proclaim that J2EE solutions are very too expensive.( Just by comparing its products with Weblogic or Websphere )
Fair Competition :)
Good news indeed,
as stated previously most of the so called j2ee development are jsp/servlet only.
The thing we really need is to have true componentized j2ee products : I want to be able to get and pay only for the things I'm gonna use and have the freedom to pick the various part of my j2ee infrastructure as and when I want.
Why the hell weblogic and websphere force you to use their own jsp, ejb containers, xml parsers and other stuffs as an over integrated package which is of course expensive when most of the time only 10% of the features are really used.
Low cost or free products (JBoss, orion) have made a step in that direction, I hope others will follow.
Agreed 100% - it is built on top of component based technology why not be able to purchase it like that? BEA+IBM = Micro$oft
I keep laughed over this one. Has money been overspent on Technology in general ? Of course !! Here we are in 2001 and on a whole people still think throwing technology at a problem is still the best way to go, so why should anyone be surprised at this headline ?
What Gartner missed, was that the whole development process is not that much different than it was 10 years ago. Sure, there have been many technology and process advances, but how many people have really learned those technologies and how to effectively use them in the development process ?
Take a long hard look at the number of factors to consider in the development of a good Internet Application. Then think about time to market, acquiring the individuals with the appropriate skills, building a realistic project plan .. yadda yadda yadda yadda. Is it any wonder that the knee-jerk reaction is for Corporate America to "throw technology" at the problem rather than effectively address the real issue ??
This discussion has turned into a beauty contest, where each participant seems to promote their favourite App-Server/Vendor.
My contribution to the discussion:
Thankyou Sun for giving us Java and J2EE, so that we can choose between operating environments and application servers!
If you have a friend working at a Microsoft shop, direct him/her to this discussion and ask if they
a. can choose vandor
b. can find a freeware alternative to their commercial .NET
c. can run their app-server on a right-sized platform
(or even are allowed to compare .NET to J2EE).
Having said that, lets go back to comparing features, price and performance!
/Vaino Vaher, Bionix AB, Sweden
Another cheaper and better J2EE application server is Inprise or Borland Application Server. Not just for Jsp/Servlets, but also for EJB deployment. CMP is the best that I've seen so far. Lots cheaper than WL, WAS, and iPlanet.
Better (or on-par) maybe... but cheaper? At $12K per CPU that still is ALOT of $$$$!
The reporters who read Gartner are acting like some industry-wide hatchet men - the kind you hire just before you fire all your good people. The "Gartnerists" run around like a pre-squashed Chicken Littles with a Mickey Mouse voice yelling "I know why the sky fell, I know why the sky fell."
Gartner is basically telling the worlds IT dept. that they've made a mistake; that Gartner knows better than they do how to architect a system and that they're stupid. Ok. People who actually have made such silly decisions are probably feeling like "oh no! Here I went and recommended IBM (or BEA) - and now Gartner says I was wrong! Gee...what was I thinking!"
And everyone else just sits there (except Peter O'Kelly) and nods their head? I don't want to get on another Dennis Miller rant here, but..
In the Gartnerist view of the world, reporters should be shagging down Lou Gershner in the hall, chasing him to his car like he was Gary Condit.. sticking a microphone in his face and asking "Is it true your company has ripped off Ford, American Express, and others for billions of dollars by telling them to use EJBs instead of JSPs?"
Get real. And give IT departments a little more credit.
Facing the Mircrosoft XP juggernaut, the high-technology server industry needs Gartner's firebrand grandstanding like Idaho needs another forest fire. Or like California needs another El Paso Energy rate hike. Well... maybe that wouldn't be so badd.... they could sell some energy back and actually make money..
But, I don't think like some people do that Bill Gates is writing all of Gartners lines. After all Garnter is positioned as a kind of know-it-all about a lot of phases of IT - not just app servers. And Gartner's reports are well written - not some Steve Balmers "I love this company!" fluff and puff pieces.
But as a whole, the technology press puts way too much faith in "groups" Gartner Group, Giga Group and so on. Why don't they instead listen to the people who we already pay to know what's going on? People like the Chairman of the Federal Reserve.
Oh, if we would only have listenned to Alan Greenspan last March the way we listen to Gartner now! If the press hadn't fallen asleep when Greenspan warned about companies "hunkering down" and "retrenchment" that "not spending is exactly the wrong thing to do to bring the economy back."
Then we wouldn't be worried about some overspending now. But no. The press was too interested in whether Greenspan is going to wack the interest rate hammer again and whether he says the "R" word that their ears close up on the rest.
A big problem in the U.S. is that access to the Internet is largely controlled by four big companies. These companies have monopoly powers and they don't have to care about things like choice. And President Bush wants to give these giants who we still call the "Baby Bells" even more power.
If Bush ever looked at what's possible, at what some countries are doing, he'd see them developing a Net that runs 130 times faster than typical American DSL and 1500 times faster than the typical U.S. home user.. But why should U.S.phone companies want to speed it up?
Why should we give them 100 MBs when we already have a monopoly and Bush to protect it? Why should we build optical fiber networks or push FTTH? We already have the entire U.S. running on copper.
In fact why should we even give them DSL? Just because we've known about it for 12 years now...
And why should we give them SIP when it means we could lose the SS7 market? And so what if optical and telco suppliers go broke? So what if we drive Corning out of business? Less competition for us. So what if Bell Labs goes broke? We don't own it anymore anyway.
No. The problem isn't overspending. The problem is called "not spending."
If the Gartnerists, like some later-day O.J. Simpson want to find "the real culprits," the IBMs and BEAs are not exactly the best places to look. It's like when you lose a quarter in the dark, searching under the street light is easier, but the quarter is still elsewhere.
But, as Mr. Miller says "that's just my opinion. And I could be wrong."
-- Rich Katz
Wow! What a rant! I have to agree that one major problem is NOT spending on IT. I think a lot of companies will pay for that later. They should spend, but spend wisely.
Companies are never very efficient, especially large ones and they DO waste hundreds of billions of dollars every single year on poorly executed tech projects. Most projects do fail, to one degree or another. The tools that one uses in a project, whether jBoss or BEA or what-have-you are only as effective as the people using them and more importantly, as effective as the project itself. Projects need executive sponsorship, realistic and focused goals, not to mention high levels of user involvment. Only after these criteria are met will the choice of tool come into play. Tools, however, don't build projects - people do. Companies obsess over standards and process (Capability Maturity Model) and have a dream of being able to replace thought, action and people with Process. They often choose development tools as a means to this particular end and therefore are certainly prone to overspending on such items.
What I hear a lot of on this list when folks praise the virtues of a large expensive tool is the fact that the vendor has tech support and that the company will 'be around' in the future or that it has such-and-such capability. These are all nice features but at the end of the day you have to ask yourself, "Yes, but will it mow the lawn?" - in other words, is it going to do the job required of it? If I buy the solid gold hammer, is it going to pound the nail better than if I buy the iron hammer?
And hey, sometimes the gold hammer IS worth having, but put some thought into it. IT departments tend to want one-size-fits-all solutions (get the one tool and we will settle on that - period). And in my experience, there is no such thing.
Tech can and will help companies succeed in the global business arena. I think halting or drastically slowing tech and IT spending is a mistake that will cost much down the road (my opinion of course!). That said, when IT departments DO spend, they should put serious thought into the purchase and not just go for the most expensive tool they can find (unless it truly fits the bill).
"replace thought, action and people with Process"
That is not my impression of what CMM or other things like it are about. I still like comparing the development process to house building. In building a house, have thoughts, actions and people been replaced by Process ? No, but they understand order.
For example, how many times have you seen a application start applying the "roof" when the foundation hasn't been layed ? How many of develop without understanding they need to release the code into a test environment ?
I could go on and on and on, but the point behind some of these efforts is to get ORDER to the process. And since most developers are never taught ORDER, exactly how do you expect them to achieve it ????
I have worked on about 4 projects (not many, I know) that were trying to achieve some level of CMM - all got caught in the Process morass and never really achieved success. I think that there are two concepts - process and Process. 'process' means a series of repeatable steps to achieve some goal and to provide some order. Design, develop, test, release - that sort of thing. 'Process' (with a cap 'P') is where HOW you do something becomes more important than WHAT you do. For instance in all CMM projects that I have worked on, our meetings stopped talking about what the project itself was trying to achieve and ignoring the project goals themselves. Instead, it was "how close are we to achieving CMM compliance?"
Software development has as its very nature some level of Chaos. We do what we can to blunt the edge somewhat by imposing some level of order (or process). But CMM and ISO 9000 (and the like) might be good for making widgets or McDonalds hamburgers or missiles the same day-in, day-out but it is much more difficult to take the inherent chaos out of software development.
Is there a neccessary order that needs imposing on software development? You bet. But just enough to achieve your goals. And I would argue strongly against wasting money on some canned Methodology.
While I have been involved with development for a while, I have learned a few things over the past couple of years. The major one being that the "chaos" you mention is in large part due to unqualified people (and I am not referring to you per-se, just a general comment) that end up in this field.
This results in most teams just not being prepared. I would argue that most development teams don't have a simple understanding of what the business is or needs. Add to that they don't have a clear/detailed business requirement, that is chaos in of itself.
But why stop there ? Even if a clear business spec existed, would they know how to create any type of application architecture, tie it into adata model, map the object model, understand load capacity etc, etc etc ??
So you are right, there is chaos in the development process. Because no one even tries to address any of the problems I just listed - its all done piecemeal and as a result its never done correctly
I agree with you.
Over my 16 year software development career, I have had the pleasure of serving on two very good teams out of god-knows-how-many. Both of those teams produced high quality software that is still in production today that the users love and use to advance their departmental and enterprise business functions. Those teams were successful not because we used or didn't use a particular development methodology or we used or didn't use a particular tool - anything like that was a means-to-an-end. Success came from something quite rare - that of 'clicking on all cylinders'.
You can't buy that and you can't make it happen. Was there chaos? Of course. But it was a creative and mentally stimulating environment that I have not worked in since and expect to only rarely. If anyone has experienced this, then you know what I mean.
I will finish this thread with this:
Tools and Methodologies do not create successful software projects, regardless of how much money one spends. They can provide a framework for sure. Companies spend an inordinate amount of money on Silver Bullets - anywhere from app servers to databases to development methodology certifications to you-name-it. None of these can compensate for poor leadership, unqualified development staff, poor or non-existent team chemistry, the inability to gather clear requirements from the user community, and the inability to execute.
By the way, if you are interested, pick up a book by Tom DeMarco and Tim Lister called "Peopleware", or one by just Tom DeMarco called "Why does Software Cost So Much?" - I think he and Tim have a lot of really good ideas on development and teams, etc. Very readable and very good!
Just to add my two cents... The almighty analysts from the Ivory Towers at Giga and Gartner are realy no different than most of us on this post. These guys have modest backgrounds, and are expressing opinion, not necessarily fact. What you choose to beleive or disbelieve from these guys is directly proportional to your trust in a specfic analyst.
One missing element in the Gartner report was to address the amount of extra capacity and scale intentionally built into a particular system. Granted most of that capacity sits idle for a good percentage of the time presenting the illusion of 'wasted' or 'overbuilt' systems. But it only takes one active day to be thankful for the extra bandwidth and capability of those systems... this is not unlike insurance for our applications.
I believe Gartner's report is about software rather than hardware. Expenses on unnecessary software, that is.
I think many people have missed the point here. The big downgrading is not on the Application server software but on the hardware it is running on. The license costs for nuch commercial software is cheaper on intel platforms than on Sun/Solaris platforms. Savings are made both on the license cost and on the hardware costs. The fact that companies have the option of downsizing their server hardware indicates one or more of the following three things:
1. J2EE is clearly at least as scalable as they originally thought, meaning that more traffic than was expected can be handled by lower spec machines.
2. They got less traffic than they expected (pretty much goes unsaid with the dot.com goes dot.bomb situation)
3. J2EE is portable so they *CAN* move to a different OS without re-writing the software. That can't be said of many other development platforms, certainly not MickeySoft (Damn I nearly did that without having a pop ;-)
I think this article is a harbinger of interesting times ahead.
J2EE doesn't need to be expensive, even if you want JMS and clustering included in the solution. There are many ~$5k/CPU app servers out there [JRUN, Lutris EAS, etc.], although they have different target audiences.
To add some context to Gartner/your comments, I recently spoke with an ISV who refused to ship their JSP/Servlet app on top of a servlet engine, even though it offered a magnitude difference in pricing to their customers.
This ISV argued that the J2EE branding campaign to CIOs/CFOs had been so successful that they ran into doors whenever they tried to pitch a servlet solution. So, this company did the amazing, they 'certified' their product only with BEA and IBM's enterprise versions, and passed on the $35k/cpu fee to their customers. In this manner, they cloaked themselves in the J2EE brand, and were able to promise the extensibility of EJB/JMS without actually having to architect their app that way. And, they were thrilled to communicate that this made it easier to license their code at $30k/cpu rather than if their code were more expensive than the underlying app server.
We're working with them to offer an affordable J2EE app server alternative to BEA, not so much that they drop BEA but rather that they offer their customers both. We'll see how it goes.
Just as a quick question without straying too much from the matter being discussed.
I've been tossing and turning over the question many people are debating over, .NET vs J2EE? Given the current economic climate and obvious shift towards cheaper solutions, how do these two behemoths stand side by side?
I have been working in both fields of recent and I'd like to be able to just turn around and pick one technology in good faith, given my background I must admit I'd love to be able to say J2EE is the way of the future, but can anyone give me a logical argument as to which technology is going to be the better of, please take into account the world is not ideal!