Enterprise HPC

Enterprise environments supporting mission critical systems like manufacturing require the highest reliability possible. The high levels of redundancy required by "6 nines" environments necessitate running with low overall resource utilization. For many fortune 500 manufacturing companies, this pool of "enterprise capacity" is vastly larger than corporate HPC resources. If we could only find a way to consume these enterprise resources for HPC while they aren't needed!

This presentation is a high level overview of a project that achieved this very result for TI – involving more than 50 people for about a year!

This is a very difficult problem from both a technology and leadership perspective. How do we consume enterprise resources with out impacting the "always on" reliability requirements? How do we maintain the security posture of each environment (IP on one side and corporate financial on the other)? How do you get "HPC people" and "Enterprise people" to work together without throwing food at each other?