Rack-a-Node to play and learn

Doesn't this look like a fun way to learn some basic cluster management concepts? Screenshot by Miriam Boon.

As the US Editor of iSGTW, I regularly search the web for the latest and greatest computing-enabled science news. A few weeks ago I was doing just that when I stumbled across something great, but quite a bit late: Purdue's Rack-A-Node game.

This game may be old news to those of you who have been frequenting SuperComputing since 2008, when it premiered on the exhibition floor. But for this editor, it was an entertaining – and productivity-killing – discovery.

The premise of the game is simple: You are the administrator of a new scientific computing cluster. With limited budgets for both money and energy, you must manage the nodes in your computer racks such that you can complete all of the science jobs that come your way before the jobs time out.

As I played the game, I found myself wondering if I was learning from the experience. To find out, I tracked down Kyle Bowen, informatics manager for Information Technology at Purdue. He told me that the game's key take-away message is the relationship between cost, space, power, and demand.

"The game is intended to replicate the decision making that goes into building a new cluster," Bowen explained. "The game attempts to balance realistic technical aspects of the cluster with having a playable game. The scope of the game would be too limited to draw a direct comparison with real-world cluster administration, but it does involve many of the same decisions."

The game, which spans 20 levels, introduces a new type of science job every few levels. Different types of jobs require different types of upgrades to your nodes, with the result that you have some pretty tough decisions to make if you want to survive the increasingly difficult levels.

"The different types of jobs do characterize the types of jobs submitted by Purdue’s faculty at the time the game was developed," Bowen said. According to Bowen, there are seven types of jobs introduced over the course of the game: chemistry, climate, rendering, pharmacy, life sciences, physics, and engineering.

You lose the game when three jobs have timed out. Bowen could not tell me if anyone has ever reached level 20, largely because the game's focus is on the score you achieve. But the highest scores have been over 100,000.

As with many games, when you lose you get the chance to immortalize yourself in their high score chart (assuming your score was high enough). But that's not the only consolation prize – the "Game over" pop-up also invites you to submit your name and contact email if you are interested in job opportunities at Purdue. It's certainly an unorthodox recruiting strategy!

When asked, Bowen said that although he is not aware of Purdue hiring anyone they discovered through the game, it has benefitted them in a more general sense.

"The game has been a popular element of Purdue’s SC booth," Bowen said. "It has certainly helped raise awareness of Purdue’s research computing capabilities."

Despite my best efforts, I've never managed to make it past the introduction of the processor-intensive physics jobs, around level 12. Some fans of the game have posted strategies for success online that supposedly will get you as far as level 15. But I haven't had a chance to try that strategy out yet, so I can't comment on how well it works.

Ready to give it a whirl? Check it out by following this link - just don't forget to drop by here to tell us how far you got, and share your strategy with the rest of the class.