on 04-25-201210:04 AM - last edited on 10-28-201311:22 PM by bcm1
My day job is focused on OpenFlow. I began ramping up on the SDN technology before last year’s launch of the Open Networking Foundation, the organization that now oversees OpenFlow standards development, among other things. OpenFlow has been a movement for years. A growing number of vendors, Brocade included, have been getting involved with OpenFlow in various ways. Standards work, development, plug-fests, sponsorships, etc. Products are appearing for sale. But with revenues still very modest, when does a movement become a market? In the case of OpenFlow, I claim the transition truly began last week, courtesy of Google’s Urs Hölzle.
Sure, there was tons of interest in OpenFlow before last week. There’s even some research-grant-funded “demand” out there. But markets need sustainable demand, and sustainable demand in the business world depends on clear, demonstrable value. That’s what changed last week. Even if we don’t know the precise details of the cost of Google’s investment, nor exactly how much they believe they have saved, Urs presented a very credible argument in support of SDN generally and OpenFlow specifically.
Are we done, then? It’s time for buyers to send out those RFQs? Watch the vendors load up the sales teams with a steep new SDN quota? Surf the financial pages for the coolest SDN Stock Index? As my grandma used to say, “Not quite hardly.”
The shift isn’t complete yet. Google showed us the value, but with a catch. They did it from scratch, built their own boxes and wrote their own code, created their own simulator, and tested and deployed the whole shebang to handle all their (internal) production DC-to-DC WAN traffic. That stunning effort is part of what makes Google preso so remarkable (especially for a firm notoriously mum about their IT architecture). But "the catch" means that the ONF and OpenFlow vendors still have plenty of work to do. In short, OpenFlow is for real, but it’s still pretty raw.
Why? One reason is that Google’s OpenFlow was not a standard version. It couldn’t be, if you look at a calendar. They were getting started when OF1.0 was a new-born, and they needed features that weren’t yet available, so they extended it. Google (one of the founding board members of the ONF) worked actively on the creation of OF1.1, but their development activity was in parallel with OF1.1, which they influenced but did not control. So they couldn’t just force the 1.1 standard to align with their internal version of OpenFlow. Nor did it make sense for them to recode their whole network that was just going live.
Another reality to recognize is that Google, as owner of both the OpenFlow controller and the OpenFlow-enabled switch, didn’t have to address the usual interoperability concerns that arise in the marketplace. Similarly, Google knew their own environment thoroughly, so they knew what testing was appropriate. Lastly, as their own customer, they could trust their “vendor”; no reference accounts were required. There was no chasm to cross.
By contrast, OpenFlow-as-a-market is only half-way there. We vendors now have a compelling case for potential value, with Google serving as a pretty effective reference account for OpenFlow. And we’ve begun the heavy lifting required for interoperability with multivendor demos and plug-fests. Yet much more work is required. Because the ONF has been (appropriately) aggressive in pushing forward with standards, the plug-fests are based on earlier versions of OpenFlow. Adoption of the most recent versions of OpenFlow will take time, and then new testing will need to be developed, etc. New workgroups have formed to bash out questions related to Configuration, and Hybrid switches and networks. The group I’m currently co-leading, the OpenFlow Future discussion group, is seeking to charter a Forwarding Plane Models workgroup with a goal of accelerating efficient adoption of new OpenFlow features onto hardware platforms while also making life easier for controllers and customers.
So it’s important to be realistic about the state of OpenFlow. Just last August, Gartner placed OpenFlow fairly low on Technology Trigger ramp of its Networking Hype Cycle, and projected 5 to 10 years until mainstream adoption. They might be right, but mainstream adoption comes long after market viability. Google’s proof-of-value will motivate other firms to adopt sooner rather than later. Look for that betting pool on who will announce next. After all, one of the key attractions of SDN is that it can bring accelerated innovation.
So make no mistake. The OpenFlow Market is coming. The train is approaching the station. Buy a ticket and get on board for an exciting ride.