Meta Group Research: J2EE Application Servers Report

Discussions

News: Meta Group Research: J2EE Application Servers Report

  1. Meta Group has released a new report on the state of J2EE application servers. There don't seem to be any earth-shattering claims.

    Excerpt

    "In terms of IT spending, J2EE application servers will remain core investments. Defining corporate application server development and deployment standards will help organizations better manage enterprisewide application delivery to increase corporate agility and lower overall application life-cycle costs by enabling server consolidation," said Daniel Sholler, vice president and director with META Group's Technology Research Services and lead author of the report. "Leading organizations are increasing developer productivity, decreasing time to market, and reducing application and technology sprawl by mandating application server standards and constraining platform proliferation.

    During the next two years, open source application servers will gain share in the J2EE market, forcing further vendor consolidation. By 2006, Linux will become the preferred platform for J2EE execution," said Michael Barnes, vice president with META Group's Technology Research Services and lead author of the report. "The J2EE application server software stack has solidified, with products becoming much more similar and migration of code between platforms becoming steadily easier."

    Key Findings

    • IBM and BEA are currently best positioned for ongoing leadership. IBM is particularly strong in terms of global presence, strategic vision, platform breadth and delivery capabilities. BEA has strength in technology leadership and a strong delivery channel.
    • Oracle is best positioned as a challenger to both IBM and BEA at the enterprise level. It has gained significant ground on the two leaders during the past 18 months, and META Group expects the gap to close further during the next six to nine months.
    • Followers in this market should be viewed as niche players. All vendors in this category will remain viable options for well-defined projects or scenarios where these vendors' specific value proposition or expertise is relevant. The common theme among all followers is a lack of presence and, in most cases, viability concerns.

    Read J2EE Application Servers Will Remain Core Investments, Says META Group

    Threaded Messages (10)

  2. Linux[ Go to top ]

    By 2006, Linux will become the preferred platform for J2EE execution," said Michael Barnes, vice president with META Group's Technology Research Services

    I wonder how's that, given Linux problems with threads. Kernel 2.6 is released only recently (AFAIK not adopted by major commercial linux distributions, though I'm not sure here), and it still has far to go to catch up with Solaris and even Windows java performance.

    Regards,
    Igor
  3. Linux[ Go to top ]

    The very same sentence caught my eye, and I started to doubt lil bit the quality of this report. Though it is hard to say, since cannot see the full text but I would hope there is some real good argumentation behind statements like this.
    Or may be it is a typical analyst "report", short on facts, high on hopes what customers of the report want to hear.
  4. Linux, Threads, etc[ Go to top ]

    We recently moved our main FTP server from a 2.4 Kernel (Debian) running Apache 1.3 to a 2.6 kernel running apache 2.0. The threading/scheduling stuff basically allowed us the same output (about 1.2-1.3 terabytes/day) with about half of the resources being used before. That's a pretty big jump, and since that is one of our core infrastructural systems, we wouldn't have done it if we felt that it wasn't stable enough.
    Fairly soon we'll be doing the same to our Java application servers. We've been keeping track of some usage statistics for the current app-servers. We'll have to see how they change after we move to the new Kernel for that stuff.
    Jason McKerr
    The Open Source Lab
    "Open Minds. Open Doors. Open Source."
  5. Linux[ Go to top ]

    Some of the major Linux distributors, including RedHat, patched the 2.6 kernel thread fixes into their 2.4 kernel distributions even before 2.6 was officially released. RedHat 9.0, for example, has the fixed Linux threads.
  6. If your using a free version of linux, one not really meant for the enterprise, then yes, the lack of POSIX-compliant threading could be a problem (assuming your using the 2.4 kernel).
    However, both SuSE 8 and RedHat have patched POSIX-compliant threading into the 2.4 kernel. Of course you must be running fairly up-to-date virtual machines (1.4.2+ I believe) to see any real speed gains (previous versions of the Sun VM assumes Linux has no POSIX threading and uses slow work arounds, IBM's VM uses some other tricks that made it a bit quicker than Sun's on Linux).
    You can expect the commercial vendors to begin integrating 2.6 into their operating systems later this year as well.
    That being said, I am running a box with the 2.6 kernel, and I am finding java to be *much* faster on Linux than Windows. The 2.6 kernel has other improvements that has jumped up it's speed by orders of magnitude in high-load situations.
    I personally think 2006 is pretty late, by the end of this year (or right now if you are experienced enough to be reliably compiling your own kernel) Linux will at least on par with everyone else, if not a tad ahead.
  7. BEA JRockit JVM recently has introduced support for Linux thin threads, which are configurable via JVM arguments. While experimental feature for now, hopefully they will make it more stable in future. For the only supported Linux distributions are RedHat AS/ES/WS 2.1, SuSE SLES 8.0 and United Linux 1.0.
  8. Linux commercial vendors as well as kernel.org guys are making Linux better every day, there's no doubt. That includes threads problem, but they really started not so long ago. Kernel 2.6 is released last December, and before that we had NPTL which wasn't really useful because java had to support it.

    So it should take some time before vendors integrate decent multithreading (that includes not only kernel, but utilities and drivers also), before they fix all bugs they implement, etc... And other OSes don't stay and wait. Solaris 10 is out there in preview, though I didn't have a good look at it yet.

    My point is that to suggest something like "in 2006 linux will rule the enterprise java", there should be sound arguments.

    Regards,
    Igor
  9. I wonder if this report matches this poll?
    http://www.manageability.org/polls/what-is-your-favorite-application-server
    It appears that web containers aren't getting the attention from the press as they deserve. 34% of the respondents prefer web containers of app servers. Is that because nobody makes money selling them?
    Carlos
  10. Marketing data vs[ Go to top ]

    This is what is in production, as per netcraft:
    http://news.netcraft.com/archives/2003/04/10/java_servlet_engines.html
    It shows that some people buy, but don't deploy.
    http://codist.typepad.com/the_codist/2004/03/j2ee_project_go.html
    Anyway, stick to Resin, Tomcat, IBM and Orion.
    .V
  11. Re. Marketing data vs[ Go to top ]

    This is what is in production, as per netcraft:http://news.netcraft.com/archives/2003/04/10/java_servlet_engines.html
    It shows that some people buy, but don't deploy.http://codist.typepad.com/the_codist/2004/03/j2ee_project_go.html
    Anyway, stick to Resin, Tomcat, IBM and Orion.
    .V
    If you read the article you will see that even the author doesn't think netcraft data is a valid indicator of app server deployments. I would hope that Meta's methodology is a bit more scientific than this.

    Commerical apps that serve web pages directly to the net from an app server are rare indeed. My interpretation of this data is that those who are unconcerned about security or performance or who are naive enough to put an app server in their DMZ are using Resin, Tomcat, IBM, and Oracle.