Oracle OpenWorld 2009: Analysis from an Application Server Perspective
17 Oct 2009 by Simon Haslam (in Events)
| After the release of Fusion Middleware 11g just over 3 months ago on 1st July several of the middleware sessions at the OpenWorld 'uber-conference' concentrated on upgrade paths and case studies of organisations in production with Middleware 11g. Mike Lehmann from Oracle presented the Application Server Roadmap talking about the current status of WebLogic (including some new fastest benchmark results) and the kinds of features to expect in the 11gR2 and 12 releases. Interestingly in that session there were about 100 delegates and, from a show of hands, it looked like they were very evenly split between OAS and WLS administrators showing, I hope, that ex-BEA administrators are now beginning to appreciate the value of participating in the Oracle community. | ![]() |
Much of Oracle's work on WebLogic since the BEA acquisition has been in capitalising on other products within the Oracle stack; for example: OEM, AD4J, TopLink, IdM and RAC integration. The latter is a particular interest area for me (see my previous presentations on Fast Connection Failover) and is an area where Oracle Internet Application Server 10gR3 was superior to WLS.
Mike talked about how Grid-Link for RAC (included in 11gR1) now allows multi-data sources to connect to RAC database services (rather than everything on an instance). It sounds as though support for the Oracle Notification Service (ONS) will be included in 11gR2 next year which will allow fast application notification, and therefore fast connection pool failover. It sounds as though we're going to have to wait for integration with the RAC load balancing advisory messages, which would allow for more sophisticated connection pool management, until Fusion Middleware version 12 though (possibly 2011).
Finally, Mike also described the focus areas for Fusion Middleware over the next few years:
- Just enough Application Server, i.e. furthering BEA's microkernel architecture and OSGi work for building a more modular, leaner java server for specific use-cases,
- High-end shared infrastructure for multi-domain management, greater levels of virtualisation and the large-scale deployment capabilities (now sometimes called as Platform as a Service) especially for shared services,
- Fusion apps foundation: enhancements for stack wide diagnostics, ease of implementation, common lifecycle management (and, I would expect, enhancements based on experiences from the first big Fusion Apps deployments).
Later in the week Oracle's Reza Shafii gave an overview of the Fusion Middleware10g to 11g upgrade options and the tooling provided so far by Oracle. This is something I'll write on in detail another day but, in summary, the complexity of upgrade is going to depend on what you are currently using IAS for:
- Web Tier: looks pretty straightforward, e.g. even the certificates in SSL wallets are migrated by the upgrade assistant.
- Identity Management: the upgrade will clearly depend on the complexity of your environment - most sophisticated ones, such as OID multi-master replication, will require more planning. Note: much of the rolling upgrades require one or more of the OID services to be running read-only during part of the process so your required application downtime will depend on whether you can tolerate that (e.g. in a reduced functionality mode).
- Forms/Reports/Portal/Discoverer: it looks like there's strong support for this, plus I would expect Forms and Reports to migrate very easily anyway due to their light use of java.
- J2EE apps: the Smart Upgrade tool connects to an online knowledge base to give you a report of affected areas (a very good idea). Some aspects will be automatically created in your new WLS environment (like data sources and deployment plans), others (such as use of some orion/oc4j specific class libraries) you will have to migrate yourself.
- SOA: reading between the lines I'd say SOA Suite, and BPEL PM in particular, has the least advanced upgrade route at the moment. This is because you cannot yet upgrade the 10g dehydration store - you have to retire existing BPEL processes first on the old 10g environment. If you have a small or pilot SOA implementation this may be acceptable (with a bit of manual work), but for a larger implementation you should almost certainly wait a few months for improvements for an in-place upgrade of the dehydration store. Note: you can upgrade your BPEL etc projects in JDeveloper now though so can start to develop your next application release on SOA 11g now.
On a more practical angle, one session had a 11g case study discussing the systems built by e-Harmony, an online dating service serving millions of customers across several countries. One interesting observation was that they didn't appear to have a web tier, such as Oracle HTTP Server/Apache or Sun, but were serving HTTP(S?) straight from the load balancers to WebLogic (though I don't know if there was a reverse proxy layer). Overall they were servicing their customer via 4 clusters and a total of 224 WebLogic Server instances. Clearly with that many moving parts automation was going to be key; they were using a combination of RPMs, Perl and WLST. Convincing evidence of their success in scripting and build management was that apparently they only have one primary middleware administrator!
On the SOA side of middleware there were quite a few interesting presentations. One slightly controversial one was by Andreas Chatziantornio from Accenture, with the great title of "The Big OESB, OSB, BPEL Cook-Off". Fellow Brit (though now defected!) Antony Reynolds was in the audience and chipped in a few useful comments when things were going a little of track. It was a full session too, showing how there's still plenty of uncertainty about the most appropriate way to integrate systems with these products.
Note that a number of the most important OOW middleware sessions will be repeated at the UKOUG conference next month - just look for the "OOW" keyword when browsing the agenda online.
That's all for now - I'll post something about the ADF news from OpenWorld once I'm back to the UK in a day or two.

