Cluster Challenge Blog
We Won!
All the hard work of the past weeks and months finally paid off: The Winner of Cluster Challenge 2008 is the ClusterMeister Team!
Last day of the competition
Everybody was finishing up and preparing for the judges to present themselves from the best side.
Val and Chris setting up the tables in front of the booth
Display are ready
The team with the advisors Nagel (red), Lumsdaine (grey)
Setting up some fancy oscillator graphics
... and in action
The judges talking to our students
And now a walk through the other teams' booths:
Purdue
Alberta
Indiana / Dresden
Taiwan
Arizona
Colorado
MIT (Stonybrook)
Last night
Things were winding down last night - some folks started playing cards while we were all waiting for those long jobs to finish. It also looked like most teams had a pretty decent experience with their systems so everybody knew pretty much, when jobs will finish.
Some impressions of the night:
The last rider of our bycicle
People passing time
Torsten on the bike to stay fit and awake
the end
No sleep, lots of fun,
Crunch the data, push the run,
Glory in the end!
So we won't officially know the results of the Cluster Challenge until tomorrow, but we're all wrapped up now. At 4pm we finished turning in our results and had a short interview with the judges (including the Man, Jack Dongarra himself!). Basically they went over what we learned, how it was interesting to work across 7 time zones, etc. This year was definitely a lot smoother overall than last year was!
Now that everything's wrapped up, no one is being secretive any longer about the results they've gotten and how many data sets they ran. It sounds like we pretty handily beat all the teams, although I'm not positive about the Taiwanese team yet. It's great, though; there's definitely been a sense of comraderie among the teams like there was last year.
I spent the night shift here last night, and that's really been the most fun time I've spent in the conference center. Card games and movies can go a long way towards entertaining sleep-deprived college students. And in the early hours of the morning it's great to find yourself discussing datasets. :)
It's been an interesting year. I have to say that I had a different kind of experience this time around. But now it's time to go laser tagging and play in an arcade, so... cheers!
Successful Monday
Monday, 8:00 a.m. - the race is on. Torsten and I are banned from the booth from now on and Jens is taking command to guide our ship through the HPCC waters. First results are right where they should be and the team is now dividing up so that the night shift (Timo and Jens) can get some sleep and are back again for the begin of the application race at 6:45 p.m.
In the meantime Val, Jupp, and Chris are discussing and optimizing our screen usage:
and Ray looks impressed if not even dazed by our performance
Things are calming down a little over the day as the HPCC score is submitted and only minor things are left to be prepared.
In the meantime I have taken refuge next to the ATM machine across the hall (there is a power outlet right next to it!)
At 6:30 Ricky Kendall announced the teams and their HPL scores (from today's run) to the present SC08 attendees:
- Purdue with a SiCortex - 694 GFLOPs
- Alberta with an SGI Altix XE cluster - 430 GFLOPs
- IU/TUD (us) with our IBM/Myricom system - 526 GFLOPs
- Tsing Hua (Taiwan) with an HP/Intel system - 703 GFLOPs
- Arizona State with an Cray CX1 and Microsoft - 404 GFLOPs
- Colorado with an nondisclosed Aspen system - 499 GFLOPs
- MIT with a Dell system - 16.8 (!) GFLOPs
At 6:45 the application race was started by the SC08 organizers and our students are doing really well. We are twittering our results as they come so stay tuned and we promise to keep you on the edges of your seats!
More impressions from setup on Saturday
Robin gave me his pictures today and I tried to condense it to a few to give everyone an impression of the setup:
Our baby now in front of our booth
Walking onto the exhibition show floor
Quiet! Engineer at work...
Releasing the docking clamps
Getting the boxes out of the rack
Almost ready to get off
An inside look
The cluster now inside the booth
Carefully laying out our fibre Myrinet cables
And connecting the nodes
This looks right!
And on the other end of the cables - the switch...
Setting up our kickass screens
Bob is trying out our attraction - the 100% green laptop recharger (also works for cell phones) --> Runs at one pedal-flop
laying waste
Too early to be
Cleaning up, but everyone
Else might as well be.
It's been an interesting day. There have been a few unfortunate incidents on the Cluster Challenge.
MIT: From what I understand, they were going to try to run things with GPUs, and they couldn't get it to work in time. So they've been filled in for by Stonybrook, the team from last year.
Colorado: Their vendor partner hasn't yet actually released the hardware they were intending to run, so there was an epic cluster rebuilding this morning.
Arizona: Running Windows Servers? It's been an interesting challenge for them. A lot of the troubles they're having are somewhat reminiscent of issues we encountered last year on OSX.
But, anyway, the time is drawing near for all our HPCC scores to be publicized. I'm anxiously awaiting our actually getting the real data sets and getting going. :) I'm even more anxious to whup up on the other teams. :P
The conference isn't as swingin' as it's going to be, but the show floor should be opening up soon. Awesome.
The night before kickoff
As to be expected some minor setbacks before we want to race, but we are staying on top of it. We also prepared the booth to look representable tomorrow (ahem, later today). Below some impressions from the activities during the day:
Robin and Jens setting things up
Torsten cycling to get his laptop charged
Scott from Myricom checking our setup
Jens leading the way :-)
Visualization running at full speed
A look into the guts of our system
Rail kit fight - since laser swords are not allowed inside the building ;-)
sexy
Decked out in sexy
Jackets and CC t-shirts,
We're all ready now!
Well, maybe not *all* ready. But close. Maria swung by today and helped us set up all the awesome posters she made (which we barely have room for in our tiny booth, incidentally :( ). The organizers came by and gave us the amazing CC08 gear that we are currently rocking. (Pictures to come.)
Things that I missed that I found in Austin (from a girl who used to live in Texas):
people say y'all
people are "fixin to" do things
cowboy hats!
TexMex
fire ants (?)
common courtesy
Anyway, we've been chewing the fat with the other teams for a while now, and it sounds like we've got an edge on them. I'm pretty excited to walk away from this competition considerably more successful than after last year's. :)
meeting
The heart of Texas,
The soft whirring of clusters,
Soon to be outdone.
That is: their clusters will be outdone by ours.
It appears that almost every team is here now. I think MIT is the only one left; they're driving. Epic.
Almost all our stuff is set up, too. Our big, intimidating visualization screens are dominating half our booth, and the bike for power generation is busily looking awesome in front of our tables. The posters have yet to be set up, but from what I've seen they'll be gorgeous and intimidating, too. :)
Additionally, I'm notorious for not reading what other people post before I post. So I apologize in advance for reposting things.
