Experiences from Building a Science Cloud with OpenStack

Accepted Session
Short Form
Scheduled: Wednesday, June 27, 2012 from 2:30 – 3:15pm in B204


How to tame your OpenStack installation for a production environment.


OpenStack is quickly becoming one of the most popular cloud computing platforms in open source. It’s been evolving rapidly to meet the demands of its commercial and academic sponsors.

The University of Oregon recently purchased a $2 million system to bring scientific cloud computing to the College of Arts and Sciences. We decided to go with OpenStack as our cloud platform, for features it had and features in the pipeline.

This talk will focus on the nitty-gritty of how we started with the stable Diablo release, moved on to the Essex development branch for added features, and hacked in our own code to integrate with the UO directory services.

I’ll cover the major components of OpenStack: Keystone, Swift, Glance, Nova, and Horizon, and how to build a working environment that goes beyond the canned DevStack distribution to include security and scalability.

Speaking experience

I am teaching a course in cloud computing this term.

My last public talk was on "Client-Side Task Support in Matlab for Concurrent Distributed Execution" at the 6th Austrian-Hungarian Workshop on Distributed and Parallel Systems (DAPSYS).

I have a talk in submission for OSCON on a similar topic, but my intention is to be more hands on for the OSBridge, getting into the dirty details of how to make OpenStack work. Here is my proposal video for it.