Tech Field Day and SDNCentral will host the Software-Defined Data Center Symposium (#SDDC13) this fall, featuring discussions of OpenFlow, software-defined networking (SDN), software-defined storage, convergence, and the greater software-defined future. Scheduled for Sept. 10 in Santa Clara, Calif., the symposium will highlight key figures from the industry and end-user community, including moderator Ivan Pepelnjak and panelist Brent Salisbury.
Ivan Pepelnjak, CCIE#1354 Emeritus, is chief technology advisor at NIL Data Communications. He has been designing and implementing large-scale service provider and enterprise networks as well as teaching and writing books about advanced technologies since 1990. He is the author of several Cisco Press books, a prolific blogger and writer, and creator of a series of successful webinars.
Panelist Brent Salisbury works as a network architect, CCIE #11972, at Kentucky State University, where he oversees various academic, research and hospital data centers. He blogs at NetworkStatic.net with a focus on disruptive technologies that have a focus on operational efficiencies.
Registration for SDDC Symposium is now open. Learn more here about how to take part.
You’ve been working hard on SDN, from OpenFlow to OpenDaylight. As a subject matter expert and key participant of many of the panels, what types of discussions are you looking forward to?
Salisbury: We have spent the past few years endlessly speculating about what can and can’t be done with SDN. I am ready to hear about integration blueprints and details of early innovators implementing production SDN networks today within datacenters.
I am also interested in SDN outside of the data center, where there is very little consensus or leadership on what direction should be taken. Leaving the comparatively limitless resources of the virtual switch for the stark reality of application-specific hardware — it will be slower than in the data center, because of longer hardware funnels.
Ironically, the lag in getting SDN capabilities in hardware exemplifies just why vertically closed network systems hamper innovation. I see open-source projects like OpenDaylight and Open vSwitch driving interoperability by having clear reference implementations for interoperability. What’s not to like about getting to work with some of the greatest network developers out there that wholeheartedly believe in what they are doing? There is a true sense of coding the future of networking.
What core architecture changes do you think are happening around the data center?
Pepelnjak: Large-scale Layer 2 forwarding and associated data-center fabrics seem to be a lost cause. Almost everyone, including the hardware switch vendors, is focusing on IP-based overlay networks supported by high-speed IP core.
The crucial questions we’re facing today are whether to implement virtual networks in endpoints (hypervisors) or in the first hardware switches (top-of-rack or end-of-row switches), and whether to use dedicated appliances or software implementations for Layer 4-7 services.
What’s your take on convergence of storage, compute and networking in the data center?
Pepelnjak: It’s a logical next step and — at least for me — a done deal. We have to sort out numerous details, but the long-term architectural directions are clear.
Do you think networking is late to the software-defined game?
Salisbury: We’re certainly later than compute, but it’s happening. I’d expect an almost identical path to the computing market’s horizontal transition. Lack of competition stifles innovation, unfortunately. Fortunately there was a handful of pioneers who wrote the right code, asked the right questions, and made timely business decisions. In any case, I think the storage guys were the last to start talking about software-defined foo anyway, so we have someone to make us feel better about ourselves.
What types of vendors do you think are the winners and losers as we transition to the software-defined data center?
Salisbury: In my experience, the explosion of virtual ports as a result of virtualization has flattened the need for a one-to-one swap from 1G to 10G data-center Ethernet. That said, we barely have SDN products, so it’s too early for me to have an opinion. It was obviously going to take a software company like VMware, post-Nicira-acquisition, to force the disruption. Fortunately, the winner in all of this will be the consumer in about every aspect. Vendors embracing community development projects in earnest are the winners, in my opinion.
Where are the new opportunities for vendors to innovate?
Pepelnjak: Automation, design standardization, and operational simplicity. We can build large-scale data-center networks with existing technologies (as clearly demonstrated by Google, Facebook, Microsoft, Rackspace, and numerous others), but we need to simplify and automate their deployment.
Ivan, who cannot afford to miss your panel at SDDC13?
Pepelnjak: Data center and cloud architects, as well as designers interested in emerging technologies, architectures, and deployment/orchestration models.
Brent, any last thoughts about SDDC?
Salisbury: Firstly, fundamentals are still the only way to cut through the marketing pitches that are coming out. The base SDN technologies are small in number, so it makes sense to understand them since the vendors are generally using similar means of developing products. Secondly, I personally think the SDDC architecture will be similarly applied in principle to the enterprise and service provider networks. It is at least a good conceptual template regardless of mechanics.