Ethics/Day-in-the-life cluster

Day One
Castaway: During an upper level math class today, multiple students used ATLAS to run compute-extensive solving algorithms on advanced linear equations. Nothing else was running on this cluster so the task was completed in the expected amount of time.

Helios: A Biology professor quickly performed a protein search using BLAST protein databases. The searches typically run quickly. BLAST is commonly used on Helios because it’s low performance risk and its relationship with the study of human health. Both of these qualities closely fit the intended use of the Helios cluster.

Day Two
Castaway: A new version of Fedora was released this morning. The cluster administrator is currently downloading the distribution and installing it on the first child node before distributing the image to the remaining nodes. No time-extensive processes are currently running and an email went out yesterday to all registered Castaway users notifying them of the expected restart later tonight.

Helios: With the Castaway cluster under maintenance, the Helios cluster is experiencing increased usage. A statistics class had planned on testing large statistical models using programs written in R, and with Castaway out of action, Helios was brought into play. The risk of creating models with R is low, and as with any other low-risk process, R can be interpreted on Helios when needed.

Day Three
Castaway: A few physics professors have been writing code in C that will use parallel processing to search for anomalies in data on ice depths in Antarctica. With their extensive amount of data, this process is expected to take anywhere from 1 to 3 days to complete. They put in a request last week for consistent uptime during this period, so extra care is taken in allowing other processes to run.

Helios: Occasionally the clusters will experience days with no use at all. Today is one of those days for Helios.

Day Four
Castaway: The math class from before has planned to run more solving algorithms on linear equations, but with the Physics processes still running in the background, less memory has been allotted for the ATLAS processes. The equations are solved in a decent amount of time without slowing down the search in ice depth data.

Helios: The cluster administrator is waiting a little while longer to observe the functionality of the new Fedora distribution on the Castaway cluster before beginning the installation on Helios. During this time, Ganglia alerts the administrator of a power failure with one of the Helios nodes. The administrator places a call to Sun’s server support center and orders a replaced power supply.

Day Five
Castaway: IIT notifies the Computer Science department of 5 new machines available for use with Castaway. The cluster administrator configures the hardware for each machine and quickly copies the master node image to each of the new nodes. Once the new nodes are up and running, they pick up some of the computation for the Physics processes that are still running.

Helios: The new version of Fedora has been successfully installed on the Castaway cluster and everything seems to be working correctly. Today the cluster administrator will repeat the process of installing Fedora on a first node before distributing the image to remaining nodes. An email notification of the expected downtime was sent to Helios users early this morning.

Day Six
Castaway: Execution of the Physics processes completes this morning and an important update to MPI is put into place. Because a new set of BLAST protein databases are expected to be released tomorrow, the administrator decides to hold off the distribution of the image to the remaining nodes.

Helios: A student research team from the Physics department at Carleton approaches the St. Olaf cluster administrator about the possibility of running their graphing model of the Milky Way on Helios. The administrator sets up the team with Helios access and they successfully install and run their code. The use of Helios over a single modern CPU saved the team nearly 5 hours of computation time.

Day Seven
Castaway: The latest BLAST protein databases are downloaded, installed, and distributed to the remaining nodes, along with the updated version of MPI.

Helios: A Sun technician is in the Northfield area today and stops by with the replacement power supply for the malfunctioning node. He quickly opens up the box, replaces the unit, and restarts the machine. The cluster continues functioning, but now with a full set of working child nodes. Later on, the administrator repeats the upgrades to MPI and BLAST on Helios.