OpenStack Summit 2013

Published: 23 Apr 2013

Last week I attended OpenStack Summit along with almost 3,000 other geeks in my favorite city, Portland, Oregon. I've been to a few conferences over the last several years but this one was especially good. Not because of the host city, the venue or the parties (although they were good too) but because of the OpenStack community.

OpenStack Summit

The NSA and OpenStack

Where else can you hear about how the NSA leverages OpenStack for their internal cloud to run "mission" apps:

Design Sessions

There were also many sessions in which the audience was integral in helping to shape the future of OpenStack by participating in design sessions. The design sessions were great for both newcomers wanting to know more about how to hack on OpenStack as well as veterans who were able to present problems and ask and answer questions. While these sessions occurred, etherpad was actively used to present issues and design questions as well as document decisions and solutions. An example is this one which supported the "How to run multiple heat engines and scaling" session. This was nice because your time wasn't monopolized by trying to take notes while trying to follow and participate in the lively discussions.

State of the Stack

One of my favorite sessions, "State of the Stack", was presented by Randy Bias, CTO of Cloudscaling in which he gave a great overview of the history and momentum of OpenStack followed by some opinionated thoughts on each major component of the OpenStack architecture. Randy also had some great slides on who the players are in the OpenStack ecosystem and what their motivations are. In closing, he claims that OpenStack is the clear winner among competitors such as Cloudstack, Eucalyptus, and vCloud. Perhaps the best illustration of this session is this slide:

state of the stack

Curvature and Donabe

Perhaps the shiniest, "ooh and ahh" session of the conference was the demo of a project by a group of University of Kent undergrad students slash Cisco engineers. Curvature, which is "an interactive visual orchestration tool for applications on OpenStack", was demoed first and illustrated a deviation from the tab and grid ui elements of Horizon, to a fluidy, floating representation of an application topology.

curvature

"Users draw their desired application topology on a canvas using a toolset of Quantum L2/L3 components and virtual machine images. This topology can then be deployed onto a running OpenStack environment at the click of a button – with Curvature handling all of the orchestration necessary for provisioning the workload, i.e. the Quantum networks and Nova VMs."

Donabe is a "recursive container service" which is loosely coupled with Curvature and can be run indendently. Although Donabe seems like a nice companion to Curvature, it seems to have a lot of overlap with the official OpenStack orchestration project, Heat.

While the source for the project is to be open sourced in the near future, they mention the backend was built using Ruby on Rails and the front end used SVG. It should be interesting to see what happens after it's released into the wild and perhaps integrated with Heat.

Common Themes

Perhaps one the the best outcomes of a conference like this is that it allows you to identify common themes across all of the OpenStack components. Naturally, one of these themes in a cloud architure is "how to scale". I saw this theme repeated in at least three different areas: Heat, Nova and Ceilometer.

In Heat design discussions, there was talk about how best to horizontally scale the "heat-engine" component as well as discussion about concurrent resource scheduling.

In Nova/Cinder/Networking, or perhaps in OpenStack as a whole, there seemed to be a need for an overall "task/workflow system". While different projects have taken different approaches based on needs at the time, there's increased potential to leverage a common service/library to be used by all. Several sessions related to "orchestration" underscores the momentum.

The newest project, Ceilometer, has the extraordinary challenge of orchestrating data measurements and aggregation in OpenStack and easily sharing that data to suit a variety of use cases including customer billing. Among the Ceilometer sessions, which spanned all four days of the summit, in "Feedback from Ceilometer users" sought to learn and improve from users' pains and successes.

Summary

To me, the summit is a testament to how well the OpenStack community...is a community...and works well together to solve common problems and create new features at a break-neck three month development cycle. Compare this with the six months (or longer) cycle of older, but not necessarily more mature competitors such as Cloudstack (which is currently a month past their last deadline). As a relative newcomer to OpenStack, I'm thrilled to have had the opportunity to attend the summit and look forward to being a part of the community and contributing to it's success.