OraTransplant

Day 2, 3, and 4 at OpenWorld

18 November 2007 at 15:21 CET | In OpenWorld 2007, Oracle |

I thought the first day of OpenWorld with my own two presentations (here and here) was a busy one. How wrong could I have been… I’ve been to a couple of Oracle conventions by now and each time you get to now a few new people. That really adds up and by now you just walk around a convention like this and meet people you know every time. This means you spend a lot of time talking to a lot of people. That’s absolutely great and one of the most important reasons why I love attending these things. The only downside is that is leaves you little or no time to blog about what’s happening.

Between all the social activities (drinks, dinners, and even a sailing trip on the bay - thanks Hylke) I even managed to find some time for attending a couple of sessions. The first one was moving beyond project level SOA. The presentation talked about the difference between “Project-Level SOA”, “Infrastructure SOA” and “Enterprise SOA”. Many organizations, including ours, start with project oriented SOA. The presentation talked about “upgrading” this approach to Enterprise level SOA. They made a number of valid points, but I guess (hope) that things will be much easier for us. Many organizations are much bigger than ours and realizing Enterprise SOA can involve many business units and multiple system architects who think they know better. Our organization is much smaller and only has a single IT department and a single system architect. Hopefully this means we don’t have some of the common pitfalls when introducing SOA into an organization.

Another presentation I attended was on DICOM Medical Imaging with Database 11g. This was a very interesting one for me. We tried image manipulation in Database 10gR2 and found it to be very very slow. The Oracle people at this presentation assured me that many (if not all) of these problems should have been gone in 11gR1. So, it looks like I really need to give the database another change and give 11gR1 a test run. The session demonstrated that the database has support for DICOM images, and you can do some neat stuff with it (extract DICOM meta data, anonymize patient information, and many more)

I also attended a presentation by Shaun O’Brian, one of (the many) JDeveloper product managers at Oracle. He had some interesting things on SOA and BPEL. He went over some troubled scenarios and asked as what could be wrong with it. It’s hard to come up with the answer in a split second, but when he explained what went wrong I sure recognized some of them. We had the exact same problem where the new “auto-load-balance” feature of AppServer 10.1.3 caused requests to the beta-test environment to end up on the production environment and vice versa.

The final presentation I attended was on Oracle Coherence. One of the many products they acquired by acquisition. This is a very intriguing product. It’s true grid computing. To put it simply, it’s sort of a RAID cluster of memory shared over many machines. There’s one “cloud” of memory shared (and duplicated) over multiple servers. The servers could even be spread out over multiple locations using a WAN for redundancy. A program, whether it is .Net, Java, or something else, can read and write information in this cloud. It’s all optimized for extreme speed and throughput. This is the underlying technology for some of the busiest web sites out there. What to think about a game betting bookmaker that receives huge volume betting on the same game, or amazon.com and the likes. The Coherence cache also features delayed write back cache. So, your application can write something to the (in-memory = fast) cache very quickly. Your program can continue and rest assured the data is safe, since it is replicated over multiple machines. Coherence will write the information to a back end data store like a database at a later time when resources allow this.

If you’re into this high-volume stuff you should really have a look at Oracle Coherence. There was even talk that this would be the internal workings of the next release of Oracle’s Enterprise Service Bus (ESB) and BPEL. This is a very fast approach to multi-server computing of ESB and BPEL. Currently, these products rely on synchronization through the database. This is inherently slow. Why not have a grid of BPEL servers that all share the same state in real time. There are nice things to come…

Two years ago I was a bit disappointed after visiting OpenWorld 2005. There was a lot of marketing hype and sales presentations. For me, it was difficult to find real interesting technical presentations. This year, I did manage to get some real interesting presentations. Also, since then I know I a lot more people from the Oracel community and this is what makes these conventions worth while. It’s the combination of technically interesting presentations and socializing and meeting up with people you already know (and get to know some new).

That’s it for now. I’m still recovering from my jet lag. Hopefully I feel much better on Monday as I’m starting driving lessons for my motorcycle then.