The above image is one from a time-lapse sequence (which we originally intended to automatically update live... but our wireless connection wasn't robust enough) taken using a Nikon 950 with a 183-degree fisheye lens mounted high above our exhibit.
This year our primary new material being presented is:
The following photos overview our exhibit. This year, our exhibit was mostly about using commodity interchangable parts for parallel processing... and the exhibit was appropriately enough built using mostly commodity parts in interesting configurations. ;-)
For nine of the ten years that we have had a research exhibit at SC, we've had a 20'x20' booth -- as we did this year. As the photo above shows, there were three major components to our exhibit: the pentagonal display, Athlon64 cluster, and the table in the rear of the booth.
The pentagonal display system was essentially the same that we had last year, but this year was topped by a simple set of 2x4" wood struts supporting a black cloth awning... which made a huge difference in the appearance of the display. Not only did the awning improve display contrast, but it also increased the apparent size and height of the display considerably. The new (borrowed) DLP projectors also have a significantly shorter throw distance, which resulted in the projected images better filling the display area. It was very effective for slide shows (using OpenOffice Impress) and running the CDR WWW interface; images with white text and black backgrounds were particularly effective. Well over a dozen people explicitly asked us what display technology we were using... still more quietly took a peek under the black leather skirt to see for themselves. ;-)
The Athlon64 cluster, which we suspect was the first Athlon64 cluster to be publically displayed (although there have been plenty of Opteron clusters before it), was kept busy running FNN design software. The 4+1 cluster's nodes were donated by AMD; the software support was Enterprise Linux donated by RedHat. The Athlon64 is the heir-apparent to the Athlon's throne as price/performance king, so even if these boxes were a bit over-the-top as cluster nodes (e.g., we call this cluster Al because of the rather excessive brushed-aluminum cases) and the price/performance on well-tuned codes does not yet match that of a good old Athlon, we're pretty sure Athlon64 is the way of the immediate future. Actually, the pricing structure probably is going to make it a pretty close race between Athlon64 and dual Opteron nodes, but everybody has been so excited about Opteron that they've barely noticed Athlon64.
The table in the rear of the booth was actually four 6' tables pushed together -- a nice trick in that it allows us to put the empty palete and boxes under the tables, so that we don't have to wait for them to come back from storage when it is time to tear-down our exhibit. On top of the tables, we had handouts, space for laptops, and a single node of KASY0. Quite a few people were interested in seeing a node of KASY0....
Rather than our tradition of having a foamboard-triangle used to mount signs for the booth, we used $13 wire shelving from Target as a frame for two backlighted signs (shown above). The signs were printed on our HP 455C E-Format printer using really cheap (semi-transparent) paper; this gave maximum brightness, but the paper had a nasty tendancy to wrinkle because it is so thin. The backlight for each is a single 26-Watt flourescent bulb, which yielded surprisingly even lighting -- although a "hot spot" is visible in the photos, it was not apparent to the naked eye.
This year, the booth staff was entirely from the University of Kentucky. From left to right: Prof. Hank Dietz, Prof. Bill Dieter, Shashi Arcot, Tim Mattox, Tamisha Thompson, and Krishna Prabhala. These folks were usually easy to identify, thanks to our exhibit shirt design... shown below (click on the image for the high-resolution version reversed for iron-on transfer):
The only thing set in stone is our name.