Discussions

News: How Developers Feel About WebSphere

  1. How Developers Feel About WebSphere (16 messages)

    Just recently an eye-opening post about time wasted with WAS and RAD got some traction on TSS forums, while also fueling more comments to the blog that sparked the thread.

    Great tools can be greatly misused. Can we verify this assertion when it comes to WebSphere? Is WebSphere being misused?

    A brand new productivity report includes some data that might be interesting for application server users, and it includes some data for IBM WebSphere usage as well. 

    IBM WAS users, 66 respondents, have reported that on average 23% of every coding hour is spent on server restarts: 69 minutes a day. Note that for this survey, it was assumed that developers spend an average of 5 hours a day coding and 3 hours a day involved in other tasks.

    If we proceed with the calculations based on the figures above, we get 345 minutes a week, and 276 hours a year lost for tedious server restarts. Assuming that the average developer gets 4 weeks of vacation, then this is already higher than 10 work-weeks of five-hour coding days lost.

    Obviously, it would be really great to get more data from IBM WAS users, as 66 is obviously not a huge survey population. If you’re an IBM WAS user and willing to contribute some data, fill in the survey to compare your experience!


    Threaded Messages (16)

  2. I'm wondering why WebSphere is selected as a primary target for complains. Did anyone tried to calculate redeploy/restart time for Oracle WebLogic? It's almost the same numbers (if not worse). And SAP WebAS 7.0 may take 5 to 20 minutes for each restart depending on number of components deployed. 

    Actually, what the research shows "between the lines" is that developers still don't adopt techniques of unit testing outside of container. Sehr schade.

  3. server restarts idle time[ Go to top ]

    Maybe I just think differently and don't have a sales agenda like ZeroTurnaround but can someone show me how they determined how much of the 345 minutes of the day wasted was idle or not. Did they monitor the brain waves of developers during this time? Did they monitor the brain waves during the rest of the day. If this is really the case then all WebSphere users should check into a clinic and we should not be concerned about what they did during those 345 minutes but what they did during the rest of the day, ;-)

  4. WebLogic startup[ Go to top ]

    Our target platform of deployment is WebLogic Server 10.3, the startup time of the server is less than a minute (there is only one application deployed on it), so I'm just confused about the things you say.

     

    By the way I agree that starting up an Oracle database takes a lot.

  5. I did not feel too bad[ Go to top ]

    Well, it is an old tradition on TSS to write some posting on how big, clumsy, heavy, etc WebSphere is.. To be fair, yes, this has some merit. I worked with IBM products and mainly with WebSphere for about 5 years(starting from version 3 to until version 6).. However, one thing I have to confess is that WebSphere was always solid(at least after fixpack 2 releases of each version).. For the last 4 years, I've been working mainly with open source products like Tomcat, Spring, Seam, JBoss, JBPM and Mule and I love them. After those tools, I never wanted to go back to WebSphere.. Saying that, I also think that once you get used to it, WebSphere is also a very good and quality product. 

    However, recently, I decided to play with Oracle SOA suite for example. To be honest, the startup time of Oracle Weblogic and the heaviness of the whole stack was the worst thing I had seen upto now. First you start the Oracle Database services in windows in 5 minutes. Then it takes 5 minutes more to start only WebLogic server...

    The startup times for JBoss or Glassfish are also not very short, although comparedly shorter than those heavyweight giants. The only exception is Tomcat, since it is only a simple servlet container.

    I guess the thing is that, once you learn and get used to tips&tricks of a product, the others always appear bad to you within those Enterprise stacks. Having used to WebSphere, I hated WebLogic and the initial problems I encountered with it.

    Finally, according to my perspective, Application Server technologies have matured a lot, the weak players in the market have already been eliminated and all existing prodcuts are quite matured. So, instead of  going to heavyweight giants, I always chose open source prodcuts(which I believe are equally well, or even better) if I have the option to. 

     

  6. I did not feel too bad[ Go to top ]

    Things have changed in the last 3 years: WebLogic and JBoss have grown in size and so in startup time. Whereas WebSphere is "stable" because it was originally large enough ;)

    First I do not understand why "server restart" is so a problem when all JEE containers are supposed to support hot deploy of EAR or WAR bundles. As I said "supposed" because in some conditions a server restart may be really necessary but it is often a trouble in the application itself.

    By the way, I have rarely used RAD+WAS and agree it is slow compared to Eclipse+JEE with a JBoss or Tomcat server. With nowadays vendor-agnostic frameworks (Hibernate, Spring...), it is really easy to have 10 efficient developers with Eclipse+JBoss/Tomcat using application hot deployment and then a QA chain for business and non-regression tests running on production target as WebSphere for instance.

    The hypothetic risk of a different behavior between WebSphere and JBoss and potential fix costs are quickly balanced thanks to the increased developers productivity.

    That is the way we successfully work in my company for JEE software development since 10 years.

     

  7. Restart is not necessary[ Go to top ]

    It is one way to workaround the issue with restart of the server, beside hotdeploying. For EJB development it is possible to link the code from a base class as long as you don't change the business interface. The server can be running a whole day if you want. When it comes to web development it can be an issue even with hot deployment, cause it dose'nt handle hot code repalcement. There are tools for that too like java rebel.

    Unitesting hence can be peformed against the business intefaces and mock the call. Or run in container tests
    using maven plugins for tomcat or jetty. It almost always enough. So the application server you use does'nt matter. It's all about the way you are used to work with your code and deployment. If using an IDE totally then maybe there can be some issues, using a modern build tool like maven solves most of these issues. Then you also be IDE agnostic by default. The IDE is good for code completion, validation and debugging and such not for everything.

     

     

     

  8. I've been using RAD and WAS for more than 5 years. WAS startup time is not a problem at all. Deployment is very quick as you save your changes. Again, as everybody knows, the startup time depends on the machine configuration also. I wonder what the configuration would be in your case to experience the the slow startup time.

  9. Great Tool?[ Go to top ]

    "Great tools can be greatly misused." Yes they can. But then again we are talking about WebSphere. A JEE server that takes an hour to install and 2GB of hard drive space. For roughly that same functionality I can get un-tar-gziping JBoss in 5 minutes and maybe 100MB.
  10. I have used tomcat/eclipse/wtp/activemq/pluto for my latest projects with maven ear builds onto Websphere. Thats because the apps I work on these days seem to be all spring / no ejbs.

  11. More a comment on the developers[ Go to top ]

    Developers shouldn't need to restart the server. If they spend that much time restarting then they are either doing it wrong and restarting every time they change something or they write really really bad code that can only be recovered from by killing the JVM.

  12. RE[ Go to top ]

    I do totally agree to the previous post. There are tool users, and there are developers. It's a big different.

  13. misspelling[ Go to top ]

    difference ist should be instead of different

  14. Well, they must be using it wrong then. I agree, WebSphere is heavy, it takes a few minutes to startup, and deploy updates. However there are a few things you can do to mitigate this.

    1. Develope all your java code and test in Tomcat container running in Eclipse.Works great for Spring/Hibernate applications

    2. Write unit tests for everything, further helps testing the java code.

    3. Deploy ear to WebSphere

    4. Ear file is exploded on the file system so you can edit the JSP code live with instant server updates. This lets you test your JSP, Javascript, CSS without a single WebSphere restart.

    Let's keep it simple and stop spreading the FUD

  15. thanks cameroon[ Go to top ]

    It was a nice article.Please post solutions..The build process need to be reduced.

  16. Websphere like most other IBM products is more like 'blow hot blow cold'.. if everything configures fine and there are no "PMR's" invovled... there might be praises.. but what is described here is common scenario.Not only WAS is many a times way behind in upgrades.. it sometimes can pose a nightmare...The sluggishness in starting and the stability are the biggest pain point...besides the cryptic configurations for which there are a trillion page of help  but nothing that can solve your problem at hand...all in all WAS has a high potential of making life of a developer frustratingly painful while impacting their productivity!! Weblogic used to be a brezze not sure what is it like after the acquisition.. JBOSS is the way to go..

  17. Find solution please[ Go to top ]

    Can anyone give solution to reduce time for the server restart and publishing?