Mike Cohen is Director of Strategic Alliances at Big Switch Networks and Sumit Naiksatam is a Member of the Technical Staff at Big Switch Networks and core developer on the Quantum project
Having just returned from a crazy week at the OpenStack Summit in San Diego, Sumit and I wanted to share our reflections, observations, and learnings over the past several days. It was an amazing (and tiring) conference but before I dive into my observations, I wanted to take a moment to welcome Sumit to the Big Switch team!
Sumit joined our team from Cisco last week right before the OpenStack Summit got underway. He has been one of the core contributors to Quantum since its early days and has played an important role in shaping the project. At Big Switch, he’s going to have ample opportunity to continue to do so with the tailwind of Open SDN behind him. His joining also demonstrates our strong commitment to the OpenStack project and community.
So, on to the summit – I believe the event represented a very positive inflection point in the evolution of OpenStack. Let me explain a few reasons why:
• The event was BIG with over 1400 people attending. As Steve Herrod noted in his talk, it very much had the feel of the first VMworld (which happened to be in San Diego as well). It was very clear that there was a critical mass of customers, vendors, and developers to create a really rich ecosystem. Folks like Rackspace, Canonical, Redhat, IBM, and HP are all making huge investments here.
• Networking, often the biggest challenge in deploying clouds, is being tackled as a first class citizen. The Quantum project has been integrated into the core of OpenStack and its sessions were standing room only. Almost every significant vendor is looking for a way to be involved. In fact, the conversation has grown beyond basic layer 2 overlays and elevated to include a broad range of network services (load balancing, firewalls, etc.) and SDN applications.
• VMware and open source: Of course, there are many ways to interpret VMware’s involvement in OpenStack. One of the most striking to me is the idea that *everyone* needs to think about open source and open standards to participate in the cloud ecosystem. Even a highly profitable player selling proprietary clouds has realized it and is joining the community. Of course, this strongly reinforces the tenets of Open SDN we’ve been promoting in the software defined networking ecosystem as well.
• People are deploying today: After finishing a panel on network virtualization, I was very excited to have a number of companies deploying both public and private clouds approach me to learn more about how SDN solutions can make their OpenStack clouds more scalable and agile.
And of course, I spent most of my time running around to a number of meetings and missed some of the most exciting discussions happening in the design summit. Luckily, Sumit has been actively involved there and wanted to share some of his takeaways from the Quantum discussions as well.
Indeed, the passion and energy around OpenStack as demonstrated at this summit was quite amazing. More specifically, the interest around network virtualization and the Quantum project was overwhelming and very gratifying. This is a true testament to the fantastic work put in by the Quantum team and is also an indicator that the community, and industry is ready to take the plunge in exploiting the promise of network virtualization, and SDN in general.
My focus during the summit was mostly around the Quantum sessions and lots of offline conversations with new and old members of this project. Here are some key points that I would like to highlight -
• Evolution of L3 capabilities – Quantum took a significant step forward in Folsom with the introduction of L3 abstractions. While functional, work still needs to be done in areas like removing gateway bottlenecks.
• Modeling the insertion of services – This has been a topic of discussion over several summits now, and some interesting proposals were presented. It’s clear that this is a hairy problem to deal with on account of the numerous deployment options that are possible, and how to effectively present a tenant facing abstraction that facilitates any of these insertion options.
• Incorporating Load Balancer capabilities – LB emerged as the clear choice of a L4-7 service that first needed attention, and could prove as a template for other services to follow in Quantum. Given the number of vendors and interest in this space, the overloaded definitions of LB capabilities and the consequent deployment options, this was an overwhelming task to say the least. As many as four sessions were devoted to this discussion, and I was struck by the thorough preparation and superb collaboration of the several vendors and contributors involved. We came out with a clear plan and a great deal of consensus on the path forward. This process should serve as a great model to other sub-projects in Quantum as well.
• Gaps in Quantum – A good amount of time was also spent in discussing support for IPv6 (and DHCP implications), DNS capabilities, basic VPN access to instances, migration of security group features from Nova, and introduction of basic tenant firewalling capabilities.
• Quantum service improvements – It was very heartening to see the emphasis on testing and the proposal of a comprehensive five-pronged strategy along with continuous integration (Tempest) gating. Other equally important aspects such as API improvements (including enhancements to the extensions’ framework), plugin libraries for shared code, and enhancements to the VIF plugging were discussing at length.
All in all, with this summit (and the prior delivery of the Folsom release), I feel Quantum successfully graduated to the mainstream and beyond the confines of basic L2 networking. Of course, there are lots of challenges, and in turn opportunities, that lie ahead. I am just as excited as the day one of the Quantum project in Diablo summit, and now probably even more as a part of the phenomenally talented Big Switch team. I am just beginning to grasp the enormous potential of the cool technology that is being developed here and looking forward every bit to bringing this to Quantum and OpenStack. Here’s to Grizzly and beyond!