OpenStack Summit Portland 2013 - Unconference Edition

Just as the Nicira NVP session by Dan Wendlandt wrapped up, I checked out the handy OpenStack Summit schedule app on my Android platform to find that none of the upcoming sessions were of particular interest - so I headed over to check out the Unconference schedule.

What is Unconference you may ask?

Unconference was an opportunity for anyone with an OpenStack-related topic they wanted to discuss to write their name and topic in a time slot on a set of easels in the main hallway.  Slots were given out to speakers on a first-come first-served basis - once the slots were all claimed the list was finalized.  I was rather delighted by the list of sessions being offered. Unconference was a smorgasbord of OpenStack-related content that had some really interesting topics throughout the summit.  In fact, Unconference was somewhat of a regular destination for me.  Any time I couldn't find something of interest in the general sessions Unconference came through for me.

I figured out that Room A104 was my next destination as I wanted to see Mirantis present on Savanna - Hadoop as a Service on OpenStack.  I walked into a packed room with standing room only, found my place over against the side wall and tuned in to hear what the Mirantis team had to say about their new project.  The back of the room filled up with other attendees and Mirantis team members smiling proudly - and rightfully so, these guys are doing some really interesting things with OpenStack.  By the time they were ready to start, the room was as packed as could possibly be.  I don't think Mirantis expected this much hoopla over Hadoop as a Service.

Before we get too far, lets just put it out there - I'm weary about calling everything 'X' as a Service on OpenStack.  Sure you can call it 'X-as-a-service' but isn't that the whole idea of addressing a pool of resources and shouldn't everything be made available as a service that's available in a service catalog?  I would have put this up as my own Unconference topic had I gotten to the easel early enough - and it probably isn't the last you'll see of this subject from me - but I digress.

The developers went through a 20 minute presentation on the features of Savanna, talked about the future road map a little bit and then they did something that defied all odds - a live demo which actually worked!  The Mirantis team was able to demonstrate how the Savanna GUI integrates fairly simply with the OpenStack Horizon dashboard as a plugin.  In the interface, the user can easily pick the base snapshot image on which to base their Hadoop deployment, type in a number of worker and tracker nodes and away OpenStack goes building out the requested Hadoop environment.

Just a few minutes later, the presenter from Mirantis logged into the tracker node and submitted a job to the cluster to calculate Pi - a job that it did with relative ease and kicked back the result in what seemed like almost no time at all.  I was fairly amazed at how quickly the Hadoop cluster was stood up with a name node, a tracker node and 5 worker nodes - from the first provisioning steps to running a job in under 10 minutes.

After the demonstration the floor was opened up to questions and comments - of which there were plenty.  Interest in Savanna was high and there seemed to be a number of developers from other projects watching this one closely.  After the tasty preview offered in this session, I'm looking forward to seeing more of what Savanna can do in future releases.  During the conference, it was announced that Hortonworks, Red Hat and Mirantis will be working on this project jointly, so I would expect to see some really amazing developments in the not-so-distant future.

Keep up-to-date on the Savanna project at http://www.savanna.mirantis.com