Welcome!

Kirk Pepperdine

Subscribe to Kirk Pepperdine: eMailAlertsEmail Alerts
Get Kirk Pepperdine via: homepageHomepage mobileMobile rssRSS facebookFacebook twitterTwitter linkedinLinkedIn


Related Topics: Java EE Journal

J2EE Journal: Article

Where Is Open Source in the App Server Surveys?

Where Is Open Source in the App Server Surveys?

Just recently Gartner reported that IBM has overtaken BEA in application server market share. The interesting thing is that Gartner's expression of market share is in a single number, dollars. While dollars are certainly an important factor in declaring a market leader, is this an accurate measure of market lead? If it is, where does that leave open source offerings such as Jonas and JBoss?

My past experiences have taught me to question claims or proclamations that one company has a lead in a particular market share. Take BEA for instance. When the application market was immature, BEA's WebLogic revenue figures included those revenues that were generated by sales of Tuxedo. Not exactly a fair measure for those companies that were only offering an application server at the time. Not that anyone expects BEA (or IBM for that matter) to be fair. After all, these companies know that people use these values to aid them in their decisions to ride with one vendor over another. As such, they understand that their survival depends upon those numbers being larger than their competitors'. When people buy into the illusion of success that is created by these numbers, they create a reality and in doing so reinforce the myth that the numbers have meaning.

The next question is, how does open source get reported? By definition, open source draws no licensing revenues and by definition carries a 0% market share. Is it just me or am I right in saying that it just feels wrong to be basing an important decision on a report that cannot accurately describe the usage of all the offerings in the space that is being considered? It leads me to question the validity of any conclusion drawn from a study that demonstrated a clear inability to adequately model the space under consideration. But then again, we are talking about perceptions and perceptions can easily become reality.

To be fair, you cannot blame Gartner for this type of reporting. After all, it is this type of information that their customers are demanding. These customers will be making large investments into systems that are critical to their business needs. They need to know if the companies they are working with will remain solvent. In addition, they need to know if these companies will remain committed to the products that they are currently trying to sell. In their minds, the ability to draw in larger-than-life revenues means that a company will remain solvent and committed.

Take HP's entry into the application server market on October 24, 2000. "We'll form the core of HP's Internet offerings," said Kevin Kilroy, Bluestone's chairman and CEO. The buzz at the time was that this purchase would quickly propel HP into third place in the application server market. By June 2002, HP was reportedly in talks to sell the technology to Oracle. This can only be read as vindication to those who ignored the hype and took notice of Bluestone's 4% market share. The reality was that no one really understood how HP could expand upon a 4% share under the economic conditions of the day. The question is, was that 4% value accurate and could the results have been different if there was a fair way to calculate market share?

It seems self-evident that if those people who were making the final purchasing decision limited themselves to only considering reports of market share, there would be no open source application servers in use. After all, who in their right mind would choose to base a business-critical application on a product that enjoys a 0% market share? Since they are in use, clearly their market share is not 0%. That said, how can you measure it?

There doesn't appear to be a good answer to the question. To the point, JBoss market share is reported to be 27%. How did BZ Media come up with the 27% figure? Again, I don't know how but here is a sample of their methodology for another study. "An e-mail was sent out to 10,000 randomly selected SD Times subscribers inviting them to participate in a study. Excluding bounces, the net number of surveys delivered was 9,792." How many of these responded is not mentioned. The fact that response is optional results in the survey being nonscientific. Sure, it's an improvement in that it appears to include open source, but is it any better or are the results any more reliable?

It's difficult to say if this survey would produce a more reliable measure of market share than revenues. What it does say is that our choice of application server almost seems random, which is what you'd expect to see in a commoditized market where you should not find any differentiating factors between competing products. If I recall correctly, this was one of the original objectives behind the J2EE.

More Stories By Kirk Pepperdine

Kirk Pepperdine has more than 10 years of experience in OO technologies. In edition to his work in the area of performance tuning, Kirk has focused on building middleware for distributed applications.

Comments (8) View Comments

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.


Most Recent Comments
David Small 06/11/04 02:12:33 PM EDT

Though not directly related ... I use WebSphere for my daytime job and JBoss for a startup that I''ve personally vested much time and some money. Without a doubt if I had opted for WebSphere on the second endeavor, the application would be way behind, perform badly, and cost much, much more. It''s ironic that the free software has proven superior in almost every aspect.

John Thompson 06/11/04 11:29:09 AM EDT

How about this ''share'' model: Poll 10,000 IT developers on the number of hours they spend weekly "working with" (loose def''n, admittedly) the various app servers. That would tell us a bunch about what''s getting hatched out NOW, versus what are the static installations resulting from choices in the past that live on for whatever purpose.

This would be a more dynamic and meaningful measure, a sort of ''weighted average'' if you will, of app server choices being made today.

Finally ... "What it does say is that our choice of application server almost seems random" ... well, not quite, because there are surely a number of app server choices not ''on the radar'' that aren''t getting their share of the random dist.

- JT, Boulder, Colorado

Dale Cosgro 06/11/04 11:25:06 AM EDT

While it''s true that Gartner and IDC report revenue, there are other vendors out there that report adoption rates. Evans Data Corporation regularly surveys corporate IT and developers and reports on adoption rates of application servers including JBoss. The adoption rate numbers are low for JBoss, but seem to be growing...

David Hibbs 06/11/04 09:26:36 AM EDT

Minor nit... "By definition, open source draws no licensing revenues and by definition carries a 0% market share."

This is not entirely correct. Source can be open and freely available yet still carry licensing revenues. Witness Linux. Redhat sells licenses, hence generating revenue, but the source is open.

Also, JBoss sells support contracts--also a source of revenue. This number is not included in the market share numbers either, and is of skewed similarly to the market share numbers because of differences in support contract styles and rates.

A more scientific market share study would be to set up an application to scan random web sites (say 10,000) and retrieve information about what server the sites are running. This of course is too easy, though, and because anyone could do it would not generate any income for Gartner via sales...

Claudia 06/11/04 03:10:59 AM EDT

Well Bob, how the survey has been done with the IT managers ? Number of companies where there is at least one licenze of JBOSS?, Number of CPU''s where the App Server were running? Consider that for some products the optimization of the code meand that to run the same application you need 10 CPU''s if use for instance Websphere and 6 if use WebLogic.... Can you clarify better?

Bob Bickel 06/03/04 08:19:55 PM EDT

IDC and Gartner base their surveys on license revenue. One of the reasons we use the BZ Resarch numbers at JBoss is because it was a survey that was completed by 1,000 IT managers. They were allowed to have multiple choices - and many customers do have multiple app server vendors in production. IBM was at 40%, BEA at 34% and JBoss at 27%. THe JBoss number was pretty impressive because the baseline from 2002 showed JBoss at 14%. You can buy the survey (which we did) if you want all the stats - including measruemetns on other markets.

Of course even those numbers can be questioned because they do not measure how many apps are on each server, or how critical those apps are, or how much services revenue was generated from them, or any number of things. What is clear by almost any measurement is that IBM and BEA are dominant and that JBoss is on a significant rise due to the new business model of Professional Open Source.

Bob Bickel
JBoss

Bob Bickel 06/03/04 08:19:52 PM EDT

IDC and Gartner base their surveys on license revenue. One of the reasons we use the BZ Resarch numbers at JBoss is because it was a survey that was completed by 1,000 IT managers. They were allowed to have multiple choices - and many customers do have multiple app server vendors in production. IBM was at 40%, BEA at 34% and JBoss at 27%. THe JBoss number was pretty impressive because the baseline from 2002 showed JBoss at 14%. You can buy the survey (which we did) if you want all the stats - including measruemetns on other markets.

Of course even those numbers can be questioned because they do not measure how many apps are on each server, or how critical those apps are, or how much services revenue was generated from them, or any number of things. What is clear by almost any measurement is that IBM and BEA are dominant and that JBoss is on a significant rise due to the new business model of Professional Open Source.

Bob Bickel
JBoss

Bob Bickel 06/03/04 08:19:49 PM EDT

IDC and Gartner base their surveys on license revenue. One of the reasons we use the BZ Resarch numbers at JBoss is because it was a survey that was completed by 1,000 IT managers. They were allowed to have multiple choices - and many customers do have multiple app server vendors in production. IBM was at 40%, BEA at 34% and JBoss at 27%. THe JBoss number was pretty impressive because the baseline from 2002 showed JBoss at 14%. You can buy the survey (which we did) if you want all the stats - including measruemetns on other markets.

Of course even those numbers can be questioned because they do not measure how many apps are on each server, or how critical those apps are, or how much services revenue was generated from them, or any number of things. What is clear by almost any measurement is that IBM and BEA are dominant and that JBoss is on a significant rise due to the new business model of Professional Open Source.

Bob Bickel
JBoss