A recent article on Barrons Online casts doubt on the validity of IDC surveys that peg IBM #2 in terms of appserver marketshare, citing a discrepancy between IDC numbers and those provided by IBM to the SEC. An analyst at Goldman Sachs is also quoted claiming that IBM can arbitrarily make its middleware revenue numbers appear larger due to hardware/software bundling deals.
The article also quotes an IDC Vice President "Where there are stats published by the vendor, our stats are reflective of those put out by the vendor."
Read IBM, IDC and the Market-Share Scramble
I am concerned about the quote by the IDC VP. Essentially, they take and publish what they are told, unless they are talking to a company like BEA where its obvious that the revenue can only come from one source (cause they are not a hardware company and do not have many unrelated products).
Can we trust any of these marketshare surveys? I think what we need is a survey by deployments/project. Perhaps TheServerSide could organize this since we are an appropriate place to poll the J2EE community. What do you guys think?
I agree, if IDC just takes the numbers, what the hell do we need IDC for? We can just call the vendors like they do.
Plus, the fact that they only seem to call the vendors is just dumb. They should get more independant information. It's amazing that something that is obviously so important in corporate stock valuations and analysis is not in any way regulated by the SEC's rule regarding full disclosure.
Just last week when IBM was discussing how they made all that money off Linux, I was wondering how they attribute that revenue. Is it bundled. If you buy a big Linux server from them, and other softare and services are included, are those attributed to Linux? Middleware? It'd be interesting to know.
Northwest Alliance for Computational Science and Engineering
I think a survey would be a great idea. But why limit it to just these these issues? Why not go the whole hog and perform "The Great J2EE Survey" - a survey on a whole load of J2EE/EJB related aspects...
I'm sure a lot of people would be very interesting to see some stats on aspects like :
1. What percentage of people developing J2EE projects are using EJB.
2. What percentage of people are using CMP entity beans / BMP entity beans / the various different O-R mapping tools.
3. JDO usage, now and anticipated future usage.
4. Aspect-Oriented Progamming usage, now and anticipated future usage.
5. Some market-share figures on the different presentation frameworks.
6. What percentage of people are developing clients as Applets / Applications / HTML / etc.
7. What percentage of people are developing applications with mixed clients.
8. What percentage of people are developing Web Services now, and what percentage are planning to develop them in the next year, two years, etc.
9. What IDE's people are using.
10. What testing tools people are using.
11. What tuning / optimisation tools people are using.
I am sure there are many, many more that would be of interest to other developers out there...
My only concern is whether it is possible to stop the results being invalidated by people with a vested interest?
Gavin, great idea. Thats actually something we're planning on doing as well. :) We'll start with your questions as a basis.
Awesome idea Gavin,
The questions as they are frased would reflect the current situation. I would also propose questions that could outline the future. E.g.
1. What IDE would you like to use ?
2. What AppServers would you recommend to others ? \
..... and so on.
That way we can see what the future trends might be. Also, you don't always use what you'd want to use.
These numbers only account for revenue. There is no way Sun has a larger share of the app server market than JBoss. We had more downloads than the whole industry sold licenses last year and that includes develop licenses. So I agree, until there is an actual poll where people are asked what they use in development and production, one can never know.
I would love to see a study such as this.
To answer Rolf's question, the number Marc states in his paper came directly from the TogetherSoft Community site in which they had a poll of Java developers to ask them what app server they were using in development. 48% of the people responded JBoss out of a total of around 2000 people that responded.
Floyd, I think we would be interested in helping to organize this even if it is not favorable to us.
I'll say up front that I'm a WebLogic fan. I've been using it since 4.51 and will continue to do so. It's not perfect, but it stands head and sholders above the alternatives out there. This is a great article that confirms something that we should already know. IBM and Oracle jump through all kinds of hoops to inflate their market share numbers.
I've been watching the JBoss guys chime in at every chance and have even read one of Fleury's papers. I've sat on my hands for as long as I can, but HAVE to call bullshit. Their claims are completely inconsistent with what I've been seeing, so I decided to look into it myself. Forget about Gartner, IDC, Meta, etc and do this simple experiment.
Search the following sites for "JBoss" and "WebLogic" to see how many jobs our out there. Here's what I see;
Monster: WebLogic 488 JBoss 38
HotJobs: WebLogic 265 JBoss 12
Careerbuilder: WebLogic 159 JBoss 11
The experiment breaks down when you try to do WebSphere because IBM calls everything in the world WebSphere, including non-Java products like MQ, so their number is inflated.
What I see is a very clear patter...a 10X+ number of WebLogic projects out there. It's a land slide folks. Marc Fleury is trying to create a perception of something that simply isn't true. I would chalk that Togethersoft poll up as an anomalous data point where Marc got all of his friends to vote.
I know where I'm continuing to invest my skills...WebLogic.
Reguardless of your religion, we it would be nice if we didn't have to put up with all of these mis-truths.
In the interest of fairness, here are Harveys Monster numbers completed as of 8pm Jan 28.
(All states, computer software, last 60 days).
There are some MQ jobs that are also looking for WAS experience just like there are Solaris admin jobs where they are looking for WebLogic admin experience. This kind of stuff inflates both sets of numbers about the same as far as I can see.
It's too time consuming to do this the hard way but a quick look on Monsters shows a very similar job mix for both products (help desk, pre sales, engs, etc).
Career builder is similar, I guess the fairest thing you could say is that the numbers for WAS and WL are comparible on the job posting sites anyway.
IBM, WebSphere Architect
bnewport at us dot ibm dot com
All views are my own and don't reflect my employer, IBM.
I agree that there is a good chunk of WebSphere out there, but let's not pretent that all thing WebSphere = J2EE...or even 21st Century. Last I checked there were 300+ products under that name.
At least we can agree on one thing...that JBoss is a very small party of the market.
Last I checked there were 300+ products under that name.
True. But most still refer to them by their more well known names (ie. MQ Series not Websphere MQ).
So who is looking for these Websphere skills? :)
The branding of many products under the WebSphere umbrella is pretty recent. I guarantee you that if people list WebSphere for job listing it is for the App Server and not the others. CICS, DB2, Lotus, and Tivoli products are not branded as WebSphere at all. I would say WebSphere Studio and WebSphere are J2EE, WebSphere Portal runs in a J2EE Server, You can do a search on MQSeries and still get job hits, MQSeries is rebranded as WebSphereMQ. If someone has a specific job for MQSeries, I guarantee it would say WebSphereMQ. If I am looking for a very specific product skill in house, why would I put just WebSphere so I can get resumes I don't like.
If you look at the descriptions of the job, I will bet the majority are for the app server.
CICS, DB2, Lotus, and Tivoli products are not branded
> as WebSphere at all
I'm not convinced that's the case. Checkout the list of
products IBM places under the WebSphere name here:
This includes CICS and also some Lotus/Tivoli products.
> IBM might be able to divide revenue up however they
> want, but why does that mean they overstate websphere
They don't overstate it because they don't state it at all. That's the point. They don't break out revenue for the individual products. IBM marketing is left to say what they want depending on the perception they want to create in a particular market.
These are the figures from UK's JobServe for the past 5 days:
ATG Dynamo: 38
Sun one: < 11 ("one Sun administrator" matches)
Borland Application Server: 1
Can you stop pointing to press releases and other bogus information from the JBoss group that continues to promote this ridicolous lie that JBoss is taking over the world? Do you believe thier download numbers? I don't. Do you believe Fleury's market share claims? I don't. The job listings are an interesting proxy for the real world use of these products. Not press releases from JBoss. Bring me another verifiable fact and I'm all ears...not these pie in the sky marketing claims.
I am pretty sure that the download numbers come
from SourceForge and are reliable enough.
Noone has ever claimed that 2 million download
equals 2 million systems or 2 million developers.
But I think it fair to say that a huge fraction of
J2EE developers has been developing on or testing
So JBoss may not have taken over the world, but
they have got the attention of the J2EE developers.
Attention from developers and a zero price tag is
a pretty good combination, so I would say that the
future of JBoss look brigth.
So if Yann's indicative search implies anything, then the result is that JBoss have 2.5% of the J2EE Server market and 1% of the market if you count in the .NET. If we should have taken into account all PHP, Perl, ASP and whatever, JBoss share could hardly be measured.
This is in spite of more free marketing and attention than any other application and (family)company in the computer field ever have received, to my knowledge. They have made a bundle from selling the documentation though..
Do you realize that the .NET numbers are sensational, stunning news?
I'm just speaking about what I see in the different companies I am (or have been) working for.
First, the guy who needs someone for doing a small VB GUI may ask for .Net competences, the one who needs somenone for a small dynamic site with 2 JSP may ask for J2EE.
Second, often a second person is in charge of finding the guy and may not understand what .Net and J2EE is. Whatever the first one wrote, he will ask it even if it's useless for the project.
I think the numbers all of you get from the job sites indicate things, are interesting but it's not everything.
Last, but not least, a survey in my country indicates that about 60% of the IT people believe their system is "better / more fancy" than the majority. Funny of course. Some people don't like to think they are not surfing on the last cool stuff. So when they ask for J2EE or .Net, it's often for a real need in 2 years, they don't use any WebLogic, WebSphere, JBoss and the like. Some companies of course have a more pragmatic approch.
Just my point of view.
"Do you realize that the .NET numbers are sensational, stunning news?"
I'm just being honest as J2EE is not the only technology for application servers of the 'server side'. These figures though should be taken with a pinch of salt as Microsoft call everything they do .NET these days. But I'm pretty sure they're not doing too bad anyway, even though this seems to come to you as a surprise. :)
As someone who has previously used Jobserve to search for J2EE contracts I would say that the figures posted above give a somewhat misleading impression as to the number of jobs/contracts advertised for .NET versus J2EE.
Whereas virtually all jobs for .NET will include .NET in their descriptions, I have found that people advertising J2EE jobs will often omit the specific app server name and will instead only mention J2EE or parts of J2EE such as EJB, Servlets, JSP, JMS, etc.
A more reasonable search to pick up all J2EE jobs would be a search string like "J2EE or EJB or servlets or JSP or JMS or weblogic or websphere or..." (inserting any other app servers you may be interested in)
For example, a search on Jobserve for
"J2EE or EJB or servlets or JSP or JMS or weblogic or websphere or jboss or jrun"
returned 785 hits
returned 498 hits
Given that .NET is a relatively new technology and so should be creating many more new job opportunities than J2EE which is relatively mature and already has vast numbers of developers employed using it, this doesn't quite sound like the "sensational, stunning news" for .NET that Rolf seems to be suggesting...
Not a lot of .NET jobs at the javajobs.com site either. My point is simply that different job sites specialize in various technologies and products and roles. Some will tend to have more Java jobs, or more WebSphere jobs, or ... maybe even more .NET jobs.
.NET will typically match all email addresses that are at a .net domain, such as "bob at techrecruiter dot net". Also, search engines often use "." as a match-all character, which means that .NET will match anything that ends in net, such as the words Internet and intranet.
So the net result (no pun intended) is that using job listing counts to figure out what is more popular is pretty close to insane. (Fitting.)
: Easily share live data across a cluster!
Unfortunately for you so can everybody go to:
search on .net and read through the job descriptions for themselves.
But maybe Jobserve too is bought by Microsoft?
Rolf, buddy, I tried to give you a balanced answer, and you go defensive on me again.
"Tierz le rideau, la farce est jouée."
: Easily share live data across a cluster!
I wonder how many of these job postings are duplicates
that are posted by recruiters who are all trying to
fill the same position?
As always, we have to be very careful about how we read and interpret numbers. A good example is the latest skills survey by Javapro:
What Java skills employers want?
As you, they are relying on keywords to evaluate what Java skills developers should have. This approach looks compelling but is actually completely flawn. Recruiters indeed tend to add this or that sexy keyword in their offer in the hope to better catch the attention. Real life example: one of my clients mentionned EJB as a required skill in a recruitment advertisement. Actually, EJB is not used, nor planned to be used at this company. Also companies not using Websphere or Weblogic will tend to write something like "prior experience with an application server like IBM Websphere or BEA Weblogic required", even if Macromedia JRun, or Pramati, or other is actually used within the company. In this later example, your count would be Websphere:1, Weblogic:1, JRun or other... 0...
- Your Java bookstore on the Web!
We are currently using WebLogic and are in the process of converting to JBoss. While numbers may currently be exaggerated, in the very near future it won't be an exaggeration.
There's really no need to pay for WebLogic. That's the simple truth. Maybe with JBoss 2 things were iffy, but I'd much rather have another fellow coworker than pay $$$ for WebLogic.
As JBoss continues to grow, BEA will get its money more from business software than from development shops. There will be a need for developers who know BEA's stuff, but I'd much rather hop on the JBoss train. The technology is so much cooler 8-)
That Togethersoft poll was about J2EE server used during development, not deployment.
I like WEBLogic too, I think it's leader in implementation of latest specs, features and documentation.
If you looked only at the job market for a Windows-to-Linux comparison or a Solaris-to-Linux comparison, you could be forgiven for thinking Linux is insignificant. But you know that the reality is totally different.
The fact is that the job market is a trailing indicator of technology trends, compounded by the fact that most recruiters and HR people are actually technology-ignorant and merely repeat buzzwords. These people are the last ones to know about new and disruptive technologies.
So sure, you can continue to invest your time and energy in Weblogic skills. I too have worked extensively with Weblogic, but I'm investing my time in learning JBoss from now on.
The inability of educated people to extrapolate a graph never ceases to amaze me...
I think from TSS's perspective, doing a survey about J2EE usage might be worth doing. It'll certainly attract a lot of attention.
But what would be the point? Who would trust the results? How could TSS make sure that this is not manipulated? How would we know that he BEA or JBoss or IBM (or .NET guys, God forbid) would just keep on clicking all day?
clicking all day ? :-)
Floyd will write that info in TSS cookie and will
allow to vote without cookies.
I dont think that they (IBM or BEA) will spend all day
deleting cookies, it's just too slow process.
will not allow to vote without cookies, of course.
That Goldman-Sach analyst must be overexposed on BEA stock or something. First he speculates "Websphere works best with IBM hardware". Go to a websphere user group and you'll probably find more deployments on Solaris and NT than AIX.
IBM might be able to divide revenue up however they want, but why does that mean they overstate websphere? Why understate DB2 revenue and lose marketshare there? Are DB2 salespeople going let DB2 sales be booked as websphere and lose commissions, bonuses etc. Not likely.
Weblogic is a good appserver but websphere has come a long way. Right now, it is the only app server to publish SPECjappserver2002 benchmark results. It deserves some respect just for that.
"According to IDC, BEA commands a 24.8% share of the Web application server market, with IBM claiming 23%. Oracle is third with 12.1% and Sun Microsystems is fourth with 7.9%."
And where is the 48% claimed by Marc Fleury in the "Why I love EJBs" article?
#"According to IDC, BEA commands a 24.8% share of the Web
#application server market, with IBM claiming 23%. Oracle
#is third with 12.1% and Sun Microsystems is fourth with
#And where is the 48% claimed by Marc Fleury in the "Why I
#love EJBs" article?
I am pretty sure that IDC count dollars and not systems.
And 500000 system x 0 dollar = 0 dollars = 0% market share.
What does it mean?
does it mean that we will expect another round of job losses?
iam really worried abt the acqs..
"I am pretty sure that IDC count dollars and not systems"
It maybe so. But it must be possible to get the correct numbers!
I have to see them or I will be dying of curiosity.
"Curiosity killed the cat"
JBoss Group Achieves Record-Breaking Two Million Downloads In 2002; JBoss Secures Position as Industry's Most Popular Java-Based Application Server
Check out the news, http://industry.java.sun.com/javanews/stories/story2/0,1072,49908,00.html
Why are you using JBoss downloads as an indication of market share ? I downloaded JBoss to learn about EJBs. So what ? Don't count my download as 'market share', people !I've never used it in a production environment and I am not likely to do so unless corporate attitudes about unsupported products change.
IBM is a leader in this market and if they get serious about J2EE and tool usability they will probably take over the top spot. Not that I think that is likely to happen ... IBM seems to be addicted to cumbersome and overly complicated workflow models. But they have grabbed big share from BEA and there is no reason to hide from that fact.
IMO the most interesting thing about the IDC report was the fact that Oracle seems to be a significant competitor. They too have gained share.
BTW - is it not ideal for developers - and for business - to have many realistic choices for technology ?
I am NOT saying that Two Million Downloads mean there will be Two Million copies of JBoss running in production. Just wanted to give you the indication that out of Two Million Downloads,some percentage of it must have made it to production.
> ... out of Two Million Downloads,some percentage of it must have made it to production
This is too funny! This is the best form to totally usless statements. How about if I assert that it is 0.0001% then it's 20 copies.
#Why are you using JBoss downloads as an indication of
#market share ? I downloaded JBoss to learn about EJBs. So
#what ? Don't count my download as 'market share', people !
#I've never used it in a production environment and I am
#not likely to do so unless corporate attitudes about
#unsupported products change.
But if your manager comes to you and say "This project
looks good, but the customer has asked us to reduce
cost of production environment - do you have any good
What will you suggest ? You know JBoss so there are a
good chance that you will suggest that !
JBoss will not take over the world tomorrow or the day
But I have no doubt that JBoss will start to sneak into
production here and there. In the start where the cost
is critical. Then they will start to choose it because
more people know it. And suddenly one day it is
We had JBoss in production. Our client kicked out a lot of money in Oracle licenses, where they could go with a cheaper solution. But I dont think, they would pay for WL or WS, they dont know even what it is.