Head to Head – JBoss v IBM, TCA

JBoss Enterprise Application Platform 5 (JBoss EAP) and IBM WebSphere Application Server 8 (WAS) are both enterprise-class application servers. This post will evaluate the total cost of acquisition (TCA) of each one based on a comparable feature set and a specific set of hardware requirements.

WebSphere is available in three primary configurations:

  • WebSphere Application Server Express (WAS Express)
  • WebSphere Application Server (WAS)
  • WebSphere Application Server Network Deployment (WAS ND)

WAS Express and WAS do NOT include session replication or advanced clustering, and are NOT meant for highly transactional applications.

Session Replication No No Yes
Advanced Clustering No No Yes
Highly Transactional Applications No No Yes

JBoss EAP includes session replication, advanced clustering, and IS meant for highly transactional applications. Therefore, WAS ND is the ONLY configuration that is comparable to JBoss EAP.

WAS ND JBoss EAP JBoss EAP Details
Session Replication Yes Yes Documentation
Advanced Clustering &
High Availability
Highly Transactional Applications Yes Yes

JBoss EAP and WAS ND both include standard enterprise-class features.

WAS ND JBoss EAP JBoss EAP Details
Java EE 5 Certified Yes Yes
Choice of Programming & Component Models Yes Yes GWT, RichFaces, Spring Framework, Apache Struts, Java EE, OSGi
Remote Administration Yes Yes Remotely Deploy Servers (Documentation) and Applications (Documentation)
Choice of OS & Database Yes Yes
Integrated Development Environment Yes Yes JBoss Developer Studio – Integrated Tooling & Runtime Components
24X7 Technical Support Yes Yes
Server Runtime
Yes Yes Modular, Extensible, Dynamic
Easy to Install Yes Yes
Centralized Management Yes Yes Manage Server & Application Configurations
Deployment Directories Yes Yes Documentation
Security Domains Yes Yes Documentation
Security Auditing Yes Yes Monitor Security Domains & Web/EJB Applications
Kerberos &
Single Sign On (SSO)

In addition, JBoss EAP and WAS ND both include advanced enterprise-class features such as intelligent load balancing while providing documentation and references.

WAS ND JBoss EAP JBoss EAP Details
Intelligent Load Balancing Yes Yes EJB, JMS
Documentation Yes Yes Link
Mission Critical Production Deployments Yes Yes

JON, included in the JBoss EAP managed subscription, is capable of SLA driven server provisioning. WAS ND is NOT.

WAS ND JBoss EAP Boss EAP Details
Elasticity with Dynamic Provisioning No Yes SLA driven server provisioning with JON.

WAS ND is NOT an open source application server.

Open Source Partial Full


Licenses v Subscriptions

Hardware Configurations

Servers Sockets / Server Cores / Processor Total Cores
Config A 4 8 8 256
Config B 4 4 8 128

Intel Xeon Processor X7560

PVU / Core (IBM) PVU (IBM) 64 Cores (Red Hat) WAS ND JBoss EAP
Config A 120 $190 $40,000 $5,836,800 $160,000
Config B 100 $190 $40,000 $2,432,000 $80,000

IBM does not sell an Intel based server that requires less than 100 PVU per core.


Management Software $0 $0
– Management Data
– Persistent Session State
$0 $0
– Database
– Application Server
$0 $0
LDAP $0 $0
Object (e.g. – POJO) Caching $0 $0
JDK $0 $25,992 /
Messaging $0 $0
HTTP Server
– Load Balancer
– Content Caching Reverse Proxy
$0 $2,500 /
(JBoss EWS)
  • JBoss EAP and WAS ND both include management and monitoring capabilities. However, WAS ND is limited to centralized management of a single cluster. JON is not. JON provides advanced management and monitoring capabilities.
  • JBoss EAP and WAS ND do NOT require a new database to be installed. Management data and persistent session state can be stored on an existing database. Since an existing database can be used, no additional database hardware is necessary.
  • JBoss EAP and WAS ND do NOT require more or less hardware than the other to be installed.
  • JBoss EAP and WAS ND do NOT require a new LDAP directory to be installed.
  • JBoss EAP and WAS ND both include support for object (e.g. – POJO) caching.
  • WAS ND includes support for a high performance messaging product. However, JBoss EAP includes high performance messaging.
  • JBoss Enterprise Web Server (JBoss EWS) and IBM HTTP server are both based on Apache HTTP Server. Apache HTTP Server can be configured as a reverse proxy that supports load balancing and content caching.

Integrated Development Environment

JBoss EAP includes 25 subscriptions with support for JBoss Developer Studio for every 16 supported cores. WAS ND does NOT include support for WebSphere for Developers.

25 Users $57,000 $0

Total Cost of Acquisition

128 Cores
256 Cores
128 Cores
256 Cores
Application Server $2,432,000 $5,836,800 $80,000 $160,000
JDK $0 $0 $25,992 $51,984
HTTP Server $0 $0 $2,500 $5,000
IDE $57,000 $57,000 $0 $0
Total $2,489,000 $5,893,800 $108,492 $216,984

Downloads: PDF

Related: JBoss EAP vs. Websphere: A TCO Analysis

Get Unstuck with JBoss


About Shane K Johnson

Technical Marketing Manager, Red Hat Inc.

View all posts by Shane K Johnson

54 Comments on “Head to Head – JBoss v IBM, TCA”

  1. Fred Muehter Says:


    Great stuff. Can you create a PDF of this? Thanks.


  2. Antonio Says:

    Hi, thank you for sharing the report.
    However some important aspect is missing, really relevant for such a big infrastructure. What is the performance of both solutions? Do we need the same cores if we’re using JBoss or WAS?



    • Shane K Johnson Says:

      The Virtuant study, here, includes performance comparisons. However, it comes to the same conclusion that I do. That neither application server will require more or less hardware than the other despite a 10% performance advantage with JBoss EAP.

      I would not prefer that an organization configure the minimum amount of hardware such that it can satisfy the performance requirements by consistently running > 50% of capacity. I would prefer that 1.5 to 2 times the minimum amount of hardware be configured. The excess capacity can be used to handle performance spikes and support modest linear growth over time. Therefore, a 10% application server performance advantage is negligible.


    • Schofield Says:

      I looked at this very thing 2 years ago with EAP 5.1 vs WAS ND 7. Found that EAP 5.1 had some synchronization issues in the arjuna code under heavy transactional scenarios. Could only get EAP 5.1 to 80% throughput of WebSphere ND 7 when running on the same hardware. However no one in their right mind would run an OLTP system at 95%-100% system resource utilization. When utilizing the hardware at 50% to handle spikes, failover of nodes, and future growth the synchronization limitations did not come into play and we found hardware could be budgeted one for one.
      I do know Redhat has been addressing the synchronization issues and at the time they indicated they would release the specj results once JBoss was limited only be the hardware when measuring throughput. I for one am really looking forward to the results for EAP 6. With the improvements in JBoss AS 7 I’m keeping my fingers crossed you will need less hardware with JBoss.


  3. michaelDRoe (@michaelDRoe) Says:

    I manage both everyday and I have to point out that with respect to TCA JBOSS may win hands down but, I’d be interested to see with TCO over a period of a few years. EAP (dispite what I read) is not easy to manage and JON’s usefullness is limited by a lack of interface so largely managing EAP is still basically manage by text file. WebSphere is not like that in fact I can manage all my servers from 1 DM and there is almost never a need to mess with a text editor. (except with props files..

    Also, This statement is false:
    JBoss EAP and WAS ND both include management and monitoring capabilities. However, WAS ND is limited to centralized management of a single cluster. JON is not. JON provides advanced management and monitoring capabilities.

    The DM can manage as many clusters as you need it to. I manage many out of one Deployment Manager.

    I want so badly to think that JBOSS is as easy to use as WebSphere because of the cost but, sadly, it just isnt.


    • Shane K Johnson Says:

      The Virtuant study is a three year TCO analysis.

      What frequently performed management tasks have you found to be not easy? By the way, have you seen the JBoss EAP 6 Beta administration console?

      I’m not sure what ‘limited by a lack of interface’ means. JON is, in fact, a web application. For example, I’ve used JON via a web browser to deploy JBoss EAP 5 instances.

      I’ve corrected the post to indicate the the management of a single cell, not a single cluster.


      • Jason Andersen Says:

        Actually, JBoss does have a multi-year, multi-customer customer case study that was published by IDC. This was a pretty rigorous analysis and the findings were very favorable to JBoss.


        We are anxious to deliver another study like this soon as this one is a little over a year old. We think the facts still stand up very well however.

      • michaelDRoe (@michaelDRoe) Says:

        A good example of what I am refering to is deploying an app to a cluster. In WebSphere I simply scope the war as im uploading it into the dmgr and click save (with a few nexts here and there) but in JON I’d have to create an ant recipe and some kind of package and apply that to servers in some fashion. Its just not clean enough that I would take the extra time to create ant scripts and packages to deploy to a couple of servers. I guess at the scale you are talking about I can understand how it would make sense. I would tend to argue at small to medium scale, Its more difficult than its worth.

        I should say I dont dislike JBOSS but, I do think that it is just not as easy as WebSphere to manage. Im hoping to get a little more from future relaeses of JON and the admin-console. I know many people that take great pride in doing every little thing by command line, VI, and script but, I am not one of those. I’m hoping that the every-day tasks get added to the consoles.

        I will check out the ver 6 admin-console.

  4. Roman Kharkovski Says:

    If there is no difference in performance, why hasn’t JBoss published any SPECj results? My experience with number of my own tests and customer tests shows that depending on the type of application, WebSphere App Server is 30% to 250% faster than JBoss. Have a look at the spec.org site – I dont see any JBoss results: http://www.spec.org/jEnterprise2010/results/jEnterprise2010.html


    • Shane K Johnson Says:

      Nice timing. Red Hat is currently preparing its SPECj submission.

      A valid performance comparison is performed in a controlled environment with a standard application while providing both the application source code and a complete hardware specification.

      The Virtuant study performed the comparison in Amazon EC2 with the open source DayTrader application, donated to Apache by IBM, and provides a complete hardware specification in the appendix.

      While performance comparisons typically result in a 10% to 30% advantage to the winner, I would say that a 250% advantage to the winner is quite unlikely and indicates a problem with the performance comparison.


      • Roman Kharkovski Says:

        Glad to hear that Red Hat is preparing SPECj result!

        As for the 250% difference – that was not average delta for entire application, but for some of the test cases:

        1 – with JMS (comparing WAS JMS vs. JBoss Messaging – pre-HornetQ – I dont have comparisons with HornetQ at this time) and also web services performance with different size and complexity of workloads.

        2 – for large and complex SOAP/HTTP messages the performance delta did go up to 250% at times. This was also on JBoss EAP 5.1, not using the Apache CXF stack that JBoss seems to be planning for EAP v6. There was a lot of tuning done on the system for both vendors.

        Once new JBoss ships with new HornetQ and CXF, I suspect these performance differences likely to decrease.

      • Shane K Johnson Says:


        1) HornetQ was provided as a technical preview with JBoss EAP 5.1.2. It is fully supported (ships with) JBoss EAP 5.1.2.
        2) JBoss WS-CXF has been fully supported (ships with) since JBoss EAP 5.1.0.

        To say that the performance differences would likely decrease is an understatement. After all, HornetQ set a new SPECjms2007 benchmark record. (link)

      • tegbird Says:

        When is the SPECjEnterprise2010 submission due for JBoss EAP.

      • Shane K Johnson Says:

        I know we’ve been working on it. I’ll find out where we are in the process.

    • Schofield Says:

      I know there are some performance results floating around using apache’s daytrader application which give a 250% nod to WebSphere. In my own study I found the application is poorly configured for JBoss. For example out of the box daytrader does not use a jms connection pool. So if you run the scenario with JMS you make a new connection to the queue every time. This can be fixed by looking up the proper JMS connection factory that does do connection pooling properly when running in a JEE container.
      I have to agree with Shane. If you see a 250% difference something is inherently flawed with the test.


      • Lee Irwin Says:

        I believe you need to correct this statement now – “JON, included in the JBoss EAP managed subscription, is capable of SLA driven server provisioning. WAS ND is NOT.”

        WAS 8.5 includes Intelligent Management, which includes SLA driven provisioning, in addition to application edition management, and application server health monitoring.

  5. Roman Kharkovski Says:

    It is interesting to see Red Hat pushing the comparison of JBoss EAP to WAS ND, however I have not see many customers using full JBoss clustering. Vast majority of JBoss installs uses standalone servers, not clustered.


    • Roman Kharkovski Says:

      WAS Base allows clustering and failover for HTTPSession data for up to 5 servers.


      • Shane K Johnson Says:

        Per IBM, WAS Express and WAS include simple failover. They do not include advanced clustering and are not intended for highly transactional applications.

    • Shane K Johnson Says:

      You can not disqualify a comparison by summarily removing a qualifying feature. If an organization is evaluating application servers from both Red Hat and IBM and plans to deploy them in a cluster, the IBM alternative to JBoss EAP is WAS ND.

      If IBM customers are deploying standalone instances of JBoss EAP as a cost effective alternative to WAS and/or to begin the migration process, I’m not surprised.


    • Rich Naszcyniec (@naszcyniec) Says:

      @roman. Can you point us to any public survey results showing a “vast majority of JBoss installs uses standalone servers, not clustered”? I would be interested in seeing that data.


  6. Roman Kharkovski Says:

    This article has a number of factual errors – for example, the cost of WebSphere for Developers is not correct. The product itself is free and the support cost of WebSphere for Developers is $290 per year. How did you come with $57K?

    The other error is that customers need to pay flat fee to Red Hat JBoss every year, while with IBM your years 2, 3, etc. are only 20% of the first year cost. Why not do 10 year comparison?

    You should compare WAS Express and WAS Base to JBoss, not WAS ND for reasons I mentioned in the other comment.

    There are other pricing errors in this document.


    • Shane K Johnson Says:

      A user license for Rational Application Developer for WebSphere Software is $2,280. The cost 25 user licenses is $57,000.

      This is a TCA analysis. It is not a TCO analysis. How is not including costs associated with the second and third year an error? That being said, 20% of the initial IBM license and support cost for 256 cores is $1,167,360. That is $1,007,360 more than the JBoss EAP ‘flat fee’ of $160,000. Per year.

      As I have previously mentioned, neither WAS Express nor WAS include advanced clustering.


      • Roman Kharkovski Says:

        Thanks for clarification.

        However comparing JBoss Studio and RAD is comparing apples and oranges. IBM has free Eclipse toolkit with free WAS for Developers with optional support for a couple of hundred dollars per year per developer – this is functionally similar to the JBoss Studio. RAD has a lot of extra features and hence the cost is higher.

        As I mentioned in my other comment, WAS Express supports HTTPSession failover for up to 2 servers and WAS Base for up to 5 servers.

      • Shane K Johnson Says:

        However, simple failover is not the same as session replication or advanced clustering.

      • Ronak Mallik (@ronak) Says:


        Can you elaborate on the “extra features” of RAD vs. plain old eclipse or “IBM … free Eclipse toolkit” (or the free, open source JBoss Tools for eclipse http://www.jboss.org/tools )

      • Roman Kharkovski Says:


        Free Eclipse toolkit for WebSphere is somewhat similar to the JBoss Studio. It provides basic tools for JEE development, testing, debugging, etc. and leverages many of the Eclipse projects (many of them contributed by IBM in the first place).

        However the Rational Application Developer (RAD) provides a whole lot more and that is why it costs money. How many of these capabilities are found in JBoss Studio: Code coverage tools? SIP? OSGi? Portlets? UML? Communication Enabled Apps (click to call, etc.)? Robust XML, XSL support (not the basic Eclipse stuff)? SAP, Siebel, JDE, etc. adapters? Profiler and runtime analysis tools? Robust JSP/JSF editor (not the basic Eclipse stuff)? Cloud deployment pattern support? SCA? and a lot more… For some people this is worth the cost and they are very willing to pay two week worth of developer salary for the tool that can shave off many weeks off the development time.

        See overview of the RAD feature set here: http://www-01.ibm.com/software/awdtools/developer/application/features/index.html?S_CMP=wspace

      • Shane K Johnson Says:

        Comparing JBoss Developer Studio to Rational Application Developer for WebSphere Software is like comparing green apples to red apples.

        In addition to Java EE 5 tooling, JBoss Developer Studio includes a visual page editor (AJAX, JSF, RichFaces, and JSP), Hibernate tooling, Seam tooling, RIA tooling, ESB tooling, business rules tooling, a data services modeler, a business process modeler, and portal tools. Tooling for CDI. Tooling for XSLT and Smooks transformations. Tooling for BPMN to JPDL conversion. Tooling for Seam / JSF portlets.


      • Shane K Johnson Says:

        Did I mention that JBoss Developer Studio includes an entitlement to Red Hat Enterprise Linux and access to Red Hat Network (RHN)?

      • Ronak Mallik (@ronak) Says:


        I believe all of the functionalities you’ve named can be found here http://marketplace.eclipse.org/ for free.

      • Roman Kharkovski Says:

        They could be found – not sure if they really offer the same function or just “sound” similar in spirit, but the point is – are those things integrated? Do they work together? Can be installed into the same Eclipse shell? Supported? Can be applied to the same project? Offer the same features as RAD? I doubt it very much.

        Same with JBoss – all the function found in JBoss EAP can be found in Community version or other Apache projects or on sourceforge.net. Why Would anybody want to pay for JBoss EAP if they can get it all from OSS community for free?

      • Shane K Johnson Says:

        1) With JBoss Developer Studio, the tooling is integrated. It does work together. It is supported.

        2) JBoss EAP is not JBoss AS.

        3) Why would anybody want to pay $3,610 per socket for 12 months of support for WebSphere Application Server Community Edition if they can get Apache Geronimo for free?

      • Roman Kharkovski Says:

        Exactly – you got my point. Yes, there is a lot of free stuff on the net, but many companies are willing to pay for the software that is tested, integrated, supported, etc., hence Rational Application Developer has a license cost.

        This is why IBM has been named the number one share holder in the worldwide application development software market according to Gartner with 24 percent of the market.

        I guess companies are voting with dollars and Rational got the most number of votes from what I can see. At least according to Gartner.

      • Shane K Johnson Says:

        I do.

        Tested. Integrated. Supported.

        Some of the very reasons why organizations purchase subscriptions for JBoss EAP and JBoss Developer Studio.

      • Rich Naszcyniec (@naszcyniec) Says:

        @roman – I got curious and decided to lookup the functional details of the IBM WebSphere Application Server Developer Tools for Eclipse, Version 8.0.4. Unfortunately there does not seem to be muc documentaiton on the subject and what I could find is the following:

        IBM® WebSphere Application Server Developer Tools for Eclipse V8.0.4 has tools for:
        -Managing the server, such as starting and stopping.
        – Starting a remote server.
        – Publishing local and server-side code.
        – Controlling incremental publishing.
        – When combined with Eclipse IDE for Java EE Developers, WebSphere Application Server Developer Tools for Eclipse provides a lightweight environment for developing Java EE applications.

        I also noticed the IBM tool does not support developing or publishing non-Java EE applications. That does not sound like much functionality added to eclipse and also a bit restrictive. Can’t you develop and deploy non-JEE applications for WAS? Perhaps the WAS for developers is full featured but the “free” develope tool certainly is not.

        JBoss Developer Studio adds far more features to Eclipse than I have mentioned for WebSphere above. PLUS, we have not even gotten into things like web application, mobile development, SOA, BPM, and other development you can also do with JBoss Developer Studio. I know you won’t be doing any SOA or BPM development with IBM WebSphere Application Server Developer Tools for Eclipse! That is saved for things like IBM Integration Developer ($) and the highly proprietary WebSphere Message Broker toolkit ($).

      • Roman Kharkovski Says:

        WebSphere Message Broker Toolkit is free for those who have Message Broker – there is no extra cost. Dont even compare that to JBoss ESB development. The tools are incomparable.

        As for BPM – I would much rather take that discussion into a separate forum where my comments and links I have posted earlier wont be moderated by Red Hat employees.

        Curious to see if you moderate the above paragraph.

    • Rich Naszcyniec (@naszcyniec) Says:

      Roman – you and I both know IBM S&S costs are not flat for a 10 year period. Unless a customer negotiates some sort of price cap for multi-year period IBM costs will go up year over year. You also mentioned basing S&S costs on first years costs. What happens if an IBM customer gets a greater discount in the first year than their entitled price as described in the Passport Advantage program? Can you comment on how that would impact year over year S&S costs for a 10 year period? As an IBM employee perhaps you can shed some light.


      • Roman Kharkovski Says:

        Rich, IBM software support does not increase year to year (like Oracle’s does). IBM software support is always (a) 20% of the PPA price or (b) customer negotiated price – whichever is lower. This makes support cost with IBM very predictable and easy to manage.

        This is in stark contrast to the Red Hat JBoss support cost, which was completely changed overnight in November 2010 when Red Hat changed its pricing from socket based to the core based. For some customers it meant that their support costs increased two fold or more – depending on the hardware. IBM does not make these kinds of changes to its pricing policy.

  7. Sreenath Venkataramanappa Says:

    We had used Jboss 4.x 5.x 7.x at big news corp, big sports news, big satellite tv in us. we never had purchased any support from jboss/Redhat, everything was community based. Well we had few issues and had a downtime for a day in a clustered environment, but we got over it.

    I had very tough time working with Websphere, but it was @ version 3,x and 4.x. Not sure now about how flexible and easy to work with.


  8. Rajesh Says:

    As a developer, JBOSS has a small footprint on the file system, in terms of structure of the application server itself. Websphere has so many artifacts , after it is installed. It just makes it difficult to understand, if i want to get down to the nuts and bolts of the system. However, i do have to agree that , websphere’s admin console is more user friendly, but at times, it does make you feel limited, by what you can do. You always wonder, if there is a better way to configure something , and if I was limited by the UI.

    Using a build and dependency management tool like Maven/ivy is lot more easier with JBOSS , compared to websphere.

    I work at a big auto insurance company , we have been using jboss for last 8 years. The code base is insanely huge , with all of our sales, claims , service , agent systems running on java based on JBOSS servers. I cant imagine using Rational Application Developer , because, it adds some much weight on top of the typical eclipse workbench, and loading a 150 project application into the workspace, will defintely slow down the workbench.

    I also hvae to mention that, I have worked on Websphere 6.x earlier , and for a small code base, it is great.


  9. tegbird Says:

    Shane ,
    I have another question regarding the JBoss Enterprise Application Platform Comparison Calculator at http://www.redhat.com/promo/eap_calculator/ . Does the cost calculation of EAP Premium include JON.


  10. Al P Says:

    I am not biased one way or the other. To be fair to WebSphere ND, there are several Intel/AMD server configurations that are less than 100 PVU, and were this way prior to this article. http://www-01.ibm.com/software/lotus/passportadvantage/pvu_licensing_for_customers.html


  11. diopba Says:



  12. Stephen Buck Says:

    It is now June of 2014 and still no JBOSS SPECj results. What happened?



  1. 2012년 3월 29일 it 기술 동향 | - March 28, 2012

    […] JBoss VS IBM TCA […]

  2. Head to Head – JBoss EAP v WebSphere ND, Total Cost of Acquisition | Brent Sordyl's blog - April 5, 2012

    […] Story: Head to Head – JBoss EAP v WebSphere ND, Total Cost of Acquisition) Like this:LikeBe the first to like this post. jboss, websphere jboss, tca, websphere High […]

  3. JBoss AS compared to WAS (by Red Hat) » gpoul's Out Of Memory Blog - April 8, 2012

    […] blog post by Red Hat’s Technical Marketing Manager Shane K Johnson titled Head to Head – JBoss v IBM seems really strange to me, and I’m not too familiar with Red Hat’s competitive JBoss […]

  4. JBoss Enterprise Application Server 5 vs. IBM WebSphere Application Server 8 | - April 12, 2012

    […] Re-Blogged from Howtojboss.com […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: