SUCCEED

   

modelingUniverse 2012
Shodor > SUCCEED > Workshops > Archive > modelingUniverse 2012

In Modeling Your Universe, Ms. Houchins started the class with a review of yesterday's disease model in AgentSheets. To get the students thinking, she questioned them about how the agents in the model interacted. Similar to how agents interact with each other in a model, supercomputers are groups of computers working together, this eventually led to today's lesson on the topic of supercomputers. When a single computer must run a long script of code, it takes forever. With cluster computers, the work can be split among all the computers. This would allow the script to run faster. Ms. Houchins then used an example, comparing the speed of one person painting a house against two people working together. After the group discussion, William brought in Shodor's portable computational cluster, Little Fe. The workshop students got the opportunity to see how Little Fe was structured and therefore had a discussion on how the close compacted structure was a benefit to the speed in which the computers may run the script.

After taking a small break, students worked together to build a supercomputer using BCCD (Bootable Cluster CD), a quick way to run parallel computing. At first they had to download the software GalaxSee; GalaxSee is a program where you can see stars change their orbit/movement (this includes speeding up and slowing down). Before starting to build a supercomputer, there must be a head node, a computer in which all jobs of the cluster are launched, to control the group. The students set up the computers so that they communicated with each other by having the head node distribute tasks to the other computers, which then returned the results to the head node. They ran into a bit of trouble when there was more than one head node, but William was able to solve the problem by shutting down any extra head nodes. With the rest of the computers, he was able to create a cluster computer.