Current Status of the "CMSOO" prototype Tracker/ECAL software at Caltech

1) We have been integrating Lucia's ooZebra into CMSOO, and this is almost complete.

2) Once done, we will begin populating a federated database on 50GByte RAID-5 disk array. This array is hosted on a Sun machine in CACR, running Objectivity v5. A second disk array is being prepared.

3) As soon as v5 is available for the Exemplar (~1 month according to Dirk Duellmann) we will move the array to that machine. Then, we will integrate part of the database into HPSS using a coupling of the non-DCE client and the Objy ams.o (subject to availability etc.).

4) The database population will initially be carried out by a process running on NT, that remotely attaches via the AMS to the Sun federation. Maximum throughput will be limited by a) 10baseT Ethernet, or b) write performance of the disk array (estimated at ~700 kBytes/sec), or c) reconstruction speed of the CMSOO code. c) is the most likely, and we don't really worry about this aspect of the tests for the moment, since the interest is to obtain a large-ish database with realistic events, rather than to maximise population performance.

5) Current timings for "raw" events show that track reconstruction for ~3000 tracker hits takes ~2 minutes on a Pentium Pro 200. ECAL clustering is very fast.

6) Once in the database, the events are being viewer by a Java applet. This applet is constructed using the beta release of Java 3D from Sun, and the JDK 1.2 beta release 3. The Java applet contains a full set of classes that match exactly those in the CMSOO application.

7) Some examples of the viewer window are shown below. Development time is very fast with this technology. Suggestions for novel ways of making use of the many features of the API (such as lighting, picking, zooming and even sound) are most welcome.

The grey shaded areas are the ECAL barrel and endcaps. The red towers are energy deposits in the ECAL. Light blue towers are reconstructed clusters. Green dots are individual hits in the Tracker. Yellow lines are reconstructed tracks. The faint blue structure visible is the Tracker geometry, taken directly from the database.

This shows a zoomed view of a different event, with a couple of spiralling tracks in the barrel. Rotation and zooming of this rather complex event is smooth on a Pentium II 233 MHz with Matrox Millenium card and 64 MBytes memory.

Julian Bunn, 15th. June 1998.