UNIVA: The Challenges Running Machine Learning Applications in Docker on Univa Grid Engine
Time:
Wednesday, June 21, 2017 03:20 pm - 03:40 pm
Room:
Booth #M-210
Breaks:
03:15 pm - 03:45 pm Coffee Break
Speaker:
Bill Bryce, UNIVA
Abstract:
Machine Learning has been around for a long time but recent advances in Machine Learning tools and frameworks makes it much easier to tackle complex multi-dimensional problems. Many of these frameworks are themselves complex and to simplify the complexity they are sometimes packaged as Docker images and can be run as Docker Containers on a machine. Running these Machine Learning containers on a Grid managed by Univa Grid Engine presented us with unique challenges in managing, running, controlling and accounting for the applications. In this presentation we will demonstrate a Machine Learning application running in a Docker container on Univa Grid Engine, discuss each issue we encountered, the solutions implemented and outstanding problems that still need to be solved.