If you are THE Objectivity-Leon, then I remember you very well. I still like to remember the dinner with you and Brian Clark (was that him?) at the RD45 workshop in Geneva 1997.

Of the two big experiments at RHIC we, the PHENIX collaboration, are still using Objectivity/DB. The STAR people have dropped its usage, because of the lack of specialists, so they claim. A mistake, if you ask me.

Objectivity is THE PHENIX database-package for calibration (that's my project) but also for online-data such as run-conditions, high voltage settings and so on. I'd guess that the schema we'll end up with will describe some 100 persistent classes. The overall size of the federation, since we are not storing raw-data, will stay below, but be perhaps close to, one TByte after two or three years of running. If you compare this to the volumes of BarBar or CERES (my old collaboration), that's small. The way in which we are using Objectivity/DB might be interesting, though.

Being a big collaboration with a dozen independent sub-systems, we had to come up with a software design which provides common interfaces and, at the same time, allows for a lot of individual freedom. We have to use several packages from the GNU-, CERN-, and commercial world, two of which are Root(CERN) and Objectivity. Since both of them have a strong tendency to make their way into every angle of your code, we decided to introduce protective layers, realized with abstract base-classes, multiple inheritance (to get object-persistency) and the use of abstract factories. I will introduce this concept with some real-world examples on the forthcoming CHEP2000 conference.

So, in conclusion, if you are looking for really big data volumes, I don't think that we are a good example. The collaboration is big, though, (some 400 people) and this is reflected also in the overall software concept, including object persistency.