High Performance Analytics

In my “day job” I have been investigating the use of high performance computing and petascale analytics as a means to understand complex networks.

Recently, I met with Dr. David Bader (Georgia Tech) and discussed its application to characterizing the relationships between the activities, facilities and labs which comprise the DoD test infrastructure.

Complex relationships are everywhere. The issue is finding a way which sorts through the muck and provides the user actionable information. With declining budget lines everyone is suddenly very interested in optimizing throughput.

Imagine the scores of DoD programs at a starting line and between that starting line and the finish line are a complex web of labs, facilities and activities whose mission it is to verify that the meets its technical requirements. The goal is get all the programs to the finish line as soon as possible AND at the lowest possible cost.

And to make it even more interesting:

* some programs get head-of-the-line privileges,
* some programs are not meeting their goals and must restart their testing program,
* some programs are severely under-funded and must “piggy-back” their testing with similar * programs to minimize cost,
* some programs are well managed (cost, schedule and performance are okay),
* some programs are poorly managed (cost, schedule and performance are awful),

It is the characterization of these complex relationships which suits the multi-variate heuristics of high performance computing. Simply put: these beasts love a complex problem.

Bottom line: High performance computing offers the opportunity to draw more information out of the available data, promising more efficiency and lower costs.