The question still remains for IBM: When is WebSphere App Server 7 coming out? The response is still the same: deer-caught-in-the-headlight stare from the IBM software reps.
Some inside information tells me a few clues about what to expect. First off, from my conversation with a certain Distinguished Engineer last year, was that WAS 7 was due sometime this year, and that it will also be built on top of Lotus Expeditor (the commercial version of Equinox framework), which is a platform for bundling OSGi modules. The rewrite of WAS 6.1 from 6.0 was no minor undertaking. In fact it is nearly rebuilt from the ground up using OSGi, and uses the JDK 5. WAS is the last of the major vendors to unleash a Java EE 5 complaint application server, but the stakes are so much higher for IBM than any other.
What IBM must get right.
There is a lot that IBM must do to have a successful next release, and its a tall order for sure. Among these are:
WAS is certainly not their flagship product, but with so many other IBM products depending on it, they've got to roll out a top-notch product. IBM is beginning to loose some market share of WAS to other vendors, while other customers are semi-patiently ready to move to new frameworks and technologies that WAS does not yet support. My hopes are up, and I'll be happy to beta test it for IBM (...hello...IBM...volunteer here...hello???).
- It MUST have a Java EE 5 compliant container. Developers are now beginning to use the JPA and some are even using EJB 3.0. The advanced features and updates for JTA are also increasingly appealing.
- They must ensure it runs flawlessly on more platforms that any other vendor. Windows, Windows 64 bit, the new AIX 6.1, the new iOS 6.1, Linux, HP-UX, Linux on Power, and of course z/OS. The z/OS port alone is a monumental undertaking in order to maximize the workload management features in the new z/10s. Then you have the new Power 6 platforms for AIX and i5/OS.
- WAS needs to run with less resources. WAS is a memory hog. Period. You need 4GB on any 64bit system in order to run it with any respectable user load. 1.5GB won't cut it like it will on Windows. In short - it needs a smaller memory footprint.
- The portlet container needs to support JSR 268. This is the latest update to the venerable JSR 168 standard.
- WAS needs an improved UI for SSL certificates. Now, some will argue with me on this point, but for the average shop that has minimal WAS expertise, setting up SSL between the HTTP plugin and from the HTTP server to the Internet is a pain in the ass.
- WAS is the underpinning for several other products including Portal, Tivoli Identity Manager, Tivoli Access Manager, Tivoli Directory Integrator, Sametime Gateway, and more. A bad WAS release will jeopardize the development of all these other products.
- Needs to run on JDK6. This has been out for over a year, and JDK 7 should be released by next year. One of the key goals of JDK 6 was improved performance. This is badly needed for a customer based entrenched on JDK 1.4 and struggling to keep up with the hardware demands (especially for WebSphere Portal!). IBM was a member of the expert group for this JDK. JDK 6 includes several improvments in the Java Management Extentions (JMX) which is used extensively in WAS for management of applications.
- WAS needs a more modular installation. If you don't want to install the EJB, SIP, or Portlet containers, then don't. It will save memory and disk space if you don't need to crank up these potentially large containers.
- IBM needs to open their Beta testing. As IBM typically does, they have a closed beta program with key customers (usually their larger ones, and those that choose to participate in early adopter programs). What IBM is doing with Jazz project (www.jazz.net) is phenomenal, and I think will revolutionize development in the SMB markets. This program encourages open participation and transparency in their development ("warts and all" as the saying goes), which in turn breeds trust in the product, and helps focus efforts on fixes and features that otherwise would have been overlooked. Imagine how many I've overlooking in this blog post.
- WAS XD which is a superseding product to ND supports controlling other application servers besides WAS, such as Tomcat, JBoss, etc. It will need to support the latest and greatest of these servers including some new entries such as Sun's Glassfish, and Oracles latest acquisition WebLogic (Oracle bought BEA recently). I don't think XD should worry about the Oracle Application server. Since BEA is now in house, and their own AS was not very widely adopted, I think it will go the way of the Beta Max tape.
- WAS will need to run other applications other than a Java EARs. Yes, that's right. Think I'm crazy? Probably, but look at the facts. Increasingly we are seeing applications run on the JVM written in other languages such as Scala, Ruby and Groovy. Sun has even made the statement that they would like to take the 'J' out of 'JVM'. However, IBM has already made inroads to this. The wsadmin scripting interface can use JACL (a java variant of TCL) or Jython (a java variant of Python). Also with the foundation of OSGi, this becomes very possible. I would love to see a .NET written application run under the control of WAS. WAS must be able to quiese and unquiese the application, provide authentication services (using NTLM/Kerberos), and be able to handle rolling updates in a clustered environment. I think they could do it, and it would be masterstroke of IBM to be able to do it. Imagine being able to leverage those .NET development skills and deploy those applications on a mainframe. Granted, it would have to be written a wee bit different, and some functionality under a .NET framework truly would not apply. But it would certainly turn the tables on Microsoft.
- WAS must support REST out of the box, along with all the other WS-x protocols. It already has the addon packs, which now can just be incorporated into the base installer. This should be dead simple easy for them.
- Finally, and most importantly they MUST market it right. They need to get preview information for the blogging community to stir up the viral aspect. They also must sell it right to be able to penetrate the market further. They will need to anticipate the Microsoft FUD, and prove ROI over other vendors (Weblogic, JBoss). They always have the open source argument which is really not an argument. IBM is a role model at creating on-ramps to their commercial software using open source counterparts (i.e. Eclipse, Geronimo, Apache, Equinox, Linux, etc). This may be the hardest task yet, however if their recent marketing attempts are any indication, I think they will do well with it.
Labels: WAS, WebSphere