I had the opportunity to speak at two Meetups in January on the work I’ve been doing for the past few months at Lithium. The first talk was at a new Meetup that Lithium started called Cloudops that is oriented towards engineers working devops roles in cloud environments. My talk was a short 15 minute talk (slides here) covering strategies for building fault tolerant systems in cloud environments and my thoughts on how to monitor those systems. My favorite part of the talk was introducing Boyd’s OODA loop as a process model for devops work. I’ve dealt a lot with the OODA loop in my past role at NIST when we were trying to improve the process for IT security incident handling in the industry. I still think OODA is one of the better process models out there since it is abstract enough to remain relevant across many disciplines while still offering value in each individual domain that binds to it.
My second talk was at the Datastax Cassandra SF Meetup hosted at Disqus. This talk (slides here) was a bit more low-level and focused on how we have been using Cassandra at Lithium for the past six months as we move to a more service-oriented architecture internally. This talk was primarily focused on our use case, data model, and all of the issues we dealt with getting Cassandra into production. I also covered the strategy we used for migrating data from MySQL to Cassandra with zero downtime. Our migration strategy was heavily influenced by a Netflix blog post covering a migration from SimpleDB to Cassandra.