UKOUG TEBS 2010 Summary for Middleware Admins
07 Dec 2010 by Simon Haslam (in Events)
A cold and snowy Birmingham ICC was the venue for the UKOUG Technology & E-Business Suite conference last week. We haven't seen weather like this so early in the winter before (or at least not whilst I've been attending the conference) and so getting there was tricky for both people living on the east of the UK and for our overseas colleagues.
When it comes to international speakers, UKOUG this year seemed to be honoured with more than ever before. Of course there are conference regulars and favourites, like Tom Kyte and Graham Wood from Oracle, as well as John King representing ODTUG. For the middleware stream we had Mike Lehmann from Canada (but essentially Oracle HQ), Frances Zhao from Oracle HQ and Doug Clarke from Canada. AMIS sent a fair contingent including Lucas Jellema and, the ever affable, Aino Andriessen. Finally, long time friend of the user group Duncan Mills was also over from Oracle HQ. I'm sure there were even more nationalities in other streams too - I met people from the US & Canada, the Netherlands (lots!), Belgium, Ireland, Austria, Denmark, Switzerland, Germany and France!
This post gives my impressions from the middleware sessions I attended and what came across as their key points:
Mike Lehmann, Oracle
Mike's teams are responsible for most of the technology I'm interested in: WebLogic and GlassFish, JRockit and HotSpot, Coherence, OHS and iPlanet, and now Exalogic. Mike is the man you'll see speaking in the most important presentations at Oracle OpenWorld and JavaOne about products this area, and so is the definitive source for information. This session described what is happening with regards to WebLogic and Glassfish, the JVMs and touched a little on Exalogic. The key messages were:
- GlassFish is the JEE reference platform so, by definition, will get features first, whereas WebLogic Server will continue as the most stable, high performance platform for typical production environments (coupled with layered products such as SOA Suite). GlassFish will continue to alternate regarding clusters - the initial GF version of a new java spec won't support clustering (for speed of release), the next main version will.
- The next major release of WebLogic Server will be JEE6 compliant. There will be a lightweight option using the new JEE6 Web Profile.
- JRockit features and instrumentation will eventually be merged with the HotSpot JVM, with the latter becoming the main development line (thus making JRockit-style features available to non-x86 platforms too).
Total Visibility into Enterprise Java Applications with JRockit Mission Control and the WebLogic Diagnostic Framework
Tomas Nilsson from Oracle in Sweden (the JRockit team) was due to give this presentation but sadly got caught up in the travel chaos. Mike stepped in with less than half an hour's notice(!), putting together some slides and finding a webcast for a demonstration. However this was obviously an area Mike felt comfortable discussing and the informal feedback I had from the session afterwards was very positive. JRockit Mission Control and Flight Recorder is a very strong product combination and it's hard not to be impressed by it. The key points were:
- Flight Recorder is a very small overhead (low %) tool running within the JRockit JVM and writing to a circular diagnostics buffer that is available for later analysis. It can be used both during normal operation (e.g. trying to hunt down a performance problem), or after a failure (perhaps heap exhaustion and restart).
- Flight Recorder is currently off by default but this could change in a future release. The intention is that it's always running so that after an issue you don't have to switch on diagnostics and then hope that the problem re-occurs.
- The WebLogic Diagnostics Framework (and accompanying console extension) is very rich and worthy of further investigation (I can see a bit of holiday work for me there!).
Guido Schmutz, Trivadis
Guido Schmutz is an Oracle ACE Director who works for Trivadis in Switzerland. I found this session very interesting and it's material presented at other events, like the recent SOA Symposium, so has been very well thought through. The key point that came across to me was that the same principles apply for SOA systems as much as traditional ones - you need to isolate areas of functionality and test, test, test! The fact that many of the processes are defined, and possibly even created, by the business users doesn't matter - complex logic and routing can still easily produce unexpected side-effects.
Mike Lehmann, Oracle
This session was actually submitted, under a rather obscure title and description, several months prior to the Exalogic launch at Oracle OpenWorld - fortunately it made it onto the agenda!
I've blogged about Exalogic previously and the situation is basically the same
as it was then - the only difference I picked up on was that Oracle has nearly finished a new Exalogic Enterprise Deployment Guide which I think will make interesting reading. On the right is a rather unimpressive photo taken on my phone of an EDG example diagram (where's Markus Eisele when you need him?!):
Paul Walker, Oracle
This session was a little bit of a surprise for me. Paul was previously a Sun employee and now works for Oracle so had a rather Sun-centric view of Identity Management (IdM). He explained how Sun had quite a few IdM products, but that they mostly worked standalone and weren't well integrated. Whilst some of Sun's products, such as one for GRC (governance, risk and compliance - a very common term in this presentation), didn't have an Oracle equivalent, there was also significant duplication. Most important for me is the overlap between Sun's Directory Server (known under quite a few previous family names, such as iPlanet and Sun ONE) and Oracle Internet Directory (OID). I have seen both at customer sites and know that Sun's product has quite a popular following, being seen as "enterprise grade" and yet more lightweight when compared to OID. It also sounds like any IdM products that were Open Source under Sun, such as OpenSSO, are likely to be/have been jettisoned.
- There are number of products that overlap in Sun and Oracle's IdM range. Most of the Sun ones are "continue and converge", really meaning EOL.
- The non-overlapping products are being gradually integrated and will fill gaps in Oracle's IdM portfolio.
- Sun's Directory Server, now called Oracle Directory Server Enterprise Edition, will remain as a strategic product for the non-Oracle-application type of environment (and has a large number of existing customers who Oracle wants to keep happy).
Frances Zhao, Oracle
"Active GridLink for RAC" is the official name for the new data source that Oracle engineering has built into WebLogic to sit alongside, and I assume ultimately replace, Multi Data Sources. This is an exciting development for me - ever since the BEA acquisition I have been looking for something from Oracle to give functionality at least comparable to what was available with the managed data source in Oracle Application Server 10.1.3. GridLink comes with far more - Fast Connection Failover (FCF) of course, but also Runtime Connection Load Balancing (RCLB) and dynamic ONS (so that you can register a set of ONS SCAN addresses, in the same way as you do for the SCAN listener in 11gR2 of the database).
Frances gave a detailed run-down of the functionality, session affinity options etc. At the moment Active GridLink for RAC will only be available on Exalogic (or as part of the Exalogic software when announced) but I'm hoping with time it will become available to all WebLogic customers. Note however that it will be available in the imminent WebLogic Server 11gR1 PS3 (10.3.4) release so you'll at least be able to try it out for yourself.
- GridLink for RAC looks significantly more robust and better integrated with the database than the old Multi Data Source, and is due to be released as part of patchset 3
- A new ONS SCAN will allow for dynamic clusters to be supported.
- You may or may not be able to use it!
This masterclass talked about upgrading from 10g to 11g, with the emphasis on in-house java applications. Frances also gave some demonstrations including some "glue code," generated by the WebLogic SmartUpgrade tool, that allows you to migrate older style JAX-RPC web services to new JAX-WS ones on WebLogic. There were discussions about cases where people have already migrated from OC4J to WLS - I would have liked more details but I suspect they were restricted by client confidentiality. One interesting comment was that many java applications aren't actually that complicated (servlets and JDBC etc) and so migrate very easily - something borne out by my own experiences too.
- Oracle has invested heavily in migration tooling, particularly WebLogic SmartUpgrade (developed over several iterations and including a knowledge-base which continues to be updated).
- The automated web service migration is intended just to get you started - for such a major change to processing for critical web services you'd want to redevelop the interface functionality using JAX-WS anyway (rather than relying on "glue code").
- SmartUpgrade inspects your application and highlights areas for investigation (and suggests anticipated complexity, and where to look in the documentation), as well as, generating some application artifacts (deployment plans etc).
Finally, I gave a couple of presentations of my own:
Unusually for me, I intended this to be a high-level, rather than technical presentation. In practice though I think I slipped into details far too often which probably contributed to me overrunning.
- WebLogic SmartUpgrade is a very useful tool - I recommend everyone runs it at least once as part of their upgrade planning.
- Less complex applications (i.e. no EJB or web services really) migrate to WebLogic without too much effort.
- You usually have more installers to run with 11g than 10g and so I would like to see the patchsets coming with a full installer option (rather than an installer you have to run on top of the base release).
- Some of the early bugs, especially around running in production mode, have been ironed out in the latest patchsets.
- Be sure to use the java node manager (and ignore some of the WLS startup advice on the internet!)
I was much more in my comfort zone with this technical presentation, having working with Cold Failover Clusters (CFC) in production since 2006 and delivered a 10g OAS/CRS version of the presentation several times in 2007. The main points were:
- Oracle supports active-passive operation under any cluster manager.
- A virtual hostname is not as easy to set up in 11g as it was in 10g AS and involves post-installation configuration.
- ASCRS may be useful for guidance and start-up scripts for HP Serviceguard, IBM HACMP and Linux HA/Heartbeat environments.
- ASCRS stills need your own filesystem management scripts (rely on mount/umount as shown in the examples at your peril!)
The most important UK Oracle conference is over again for another year. After 2009's very significant middleware launches I had expected this year to be a little quieter, but that wasn't the case - if anything we are only now working out how best to use all this new technology. Coupled with Oracle's progress on the Sun integration, and the developments in enterprise Java, I'll say it again... we're living in interesting times!