The Day The Cluster Was Built

With a target operational date of April 15, 2000, we had a number of delays in obtaining the parts for KLAT2. Finally, the last of the parts were shipped, and we scheduled the cluster to be built on April 11, 2000. To attract students to help with the construction, we made and distributed the following poster:

Our poster is based on one of the posters for the classic 1951 science fiction movie The Day The Earth Stood Still. Yes, KLAT2 is an obscure reference to Klaatu, the fellow from outer space who came, with the robot Gort, to explain to all the people of the earth that if humans cannot work together in peace, the earth will be destroyed for the good of all planets. Of course, in the movie, Gort didn't have an AMD Athlon logo on his chest and Tux, the Linux penguin, wasn't the actor inside the suit... it's a very good movie anyway. ;-)

Over 30 students, mostly from the University of Kentucky Electrical Engineering and Computer Science Departments, came to work on the cluster. Things went very smoothly, especially given that most of the students had never seen the inside of a PC before. Although construction really began at about 11AM, all 66 PCs were built, individually booted, and placed on the shelves before 10PM. The bad news is that a shipment containing many of our network cables did not arrive on schedule, so we had to delay wiring the network. KLAT2's first "normal" boot was on our target date: April 15, 2000. However, things dragged on a bit because we had trouble configuring some software to understand our network structure and, worse still, we had a few intermittently-bad NICs and switch ports.

From April 24, 2000 through mid-May, KLAT2 was stable and no new hardware or software problems surfaced. Then we quickly lost 6 of 10 switches! In case you hadn't guessed, they have a design flaw. The manufacturer has confirmed and corrected the design flaw (well, that's what they tell us) and replacements for all 10 switches are on their way at no cost. We're still not too happy because it has taken a while to get the problem corrected. More on this later....

Independent of that, KLAT2 still is not in its final form. We plan to be augmenting each of the PCs with a parallel content-addressible memory (CAM) system that has a MIPS processor and two CAM engines on a single PCI card. KLAT2's flat neighborhood network probably will be re-wired to a better flat-neighborhood design as we refine the genetic search program that designed it; also, a second custom-built network (based on our old PAPERS designs) will be added to accelerate aggregate function communication. We might also upgrade the PCs with more memory, disks, and/or faster Athlons. Each of these changes will enhance KLAT2's abilities, but the basic functionality of KLAT2 as it is now will continue to be supported.

The Aggregate. The only thing set in stone is our name.