If you’re subscribed to my RSS feed (hi, mom!) you might already have seen these pictures. Nonetheless, I’ll add them here with a little explanation. First, the front of the cluster:
I suggest that you click on it and go to the Flickr page, as I’ve added some notes to the various parts of the picture, pointing out the major pieces. Here’s the back:
As the title of the post states, it’s supposed to arrive here in Baltimore, from San Francisco, sometime tomorrow. I’m guessing in the afternoon. A couple of guys from Penguin will be here at 13:00. That’s the good news. Now to the bad news — the news that kept me from falling asleep for hours on Friday night.
It won’t be ready to go when it arrives.
No, contrary to all of the lovely visions I was given about the cluster rolling out of the crates, and being ready to go, it will not be ready to go when it gets here. This is in large part due to all of the delays on the part of AMD, and their apparent inability to ship anything when they say they will. Now, I’m not sure just how un-ready it will be. If I can get all of the wires hooked up, and the Infiniband MPI libraries are ready to go, I should be able to have simulations running on it by Tuesday. Penguin’s software people are not excited about this, and want me to wait. We’ll see how things look tomorrow.
November and December are our deadline season, with Heart Rhythm abstracts due on January 4th. Now more than ever, we have a huge backlog of simulations to run on large models, requiring the high-performance power that this cluster packs. However, the complete set-up is not due to start until November 28th. That is no good. To make matters worse, our favorite TeraGrid cluster in Texas, Lonestar, is on the lam (no pun intended) until further notice.
I’m not superstitious, but if I were, I’d be crossing my fingers right now.