Linux Cluster Management System
(2005 – 2006)

Background

This was a cooperative academic project between Seattle University's Master of Software Engineering (MSE) program and Boeing Integrated Defense Systems (IDS).

Task

Boeing IDS had Linux-based supercomputing clusters that were used by many different people to run different types of simulations. Managing these clusters was difficult and the original methods did not allow for subdividing cluster resources, efficiently scheduling or prioritizing jobs, or estimating finish times. Our goal was to address as many of these problems as possible within an academic year.

Approach

I worked with two other graduate students at Seattle University, and a small team from Boeing IDS, to solicit requirements and desired features from potential users of the system. We analyzed these features to design a flexible architecture based on distributed Ruby. We configured our own miniature simulation cluster at Seattle University and developed the base architecture with a minimal set of features. From that point, we met with Boeing on a regular basis to iteratively solicit requirements and develop additional functionality.

Technologies Used

FreeBSD, GTK, Linux, Ruby/DRB, Subversion

Deliverables

• Source code and documentation
• Final project report
• Presentation at the annual MSE Project Night for Seattle University