Brocade is holding its analyst day today (September 12, 2012) and there are live-streaming sessions about Brocade’s vision for fabrics, software-defined networking (SDN) and clouds. Given that Brocade made announcements around SDN and OpenFlow support earlier this year (1G, 10G and yes, 100G), we felt it would be educational to sit down with Senior Director of Product Management, Keith Stewart, at Brocade to understand their position on SDN.
SDNCentral: Brocade is known for SANs (storage area networks). Can you provide a brief description of non-storage networking at Brocade? What product categories does Brocade play in?
Keith: “Brocade is a broad data center networking provider. In addition to our SAN portfolio, Brocade has a family of Ethernet Fabric products for data centers, carrier-grade routers for DC edge and WAN backbone, and high performance ADCs (application delivery controller) for load balancing.”
SDNCentral: What is your view on SDN? What does SDN change? Where is SDN a good fit? Where is SDN not a great fit?
Keith: “SDN has the potential to be an inflection point in networking, transforming the way we’ve built and operated networks for decades. At Brocade, our SDN strategy is composed of three distinct elements:
i) Network Virtualization, or the ability to create arbitrary L2 segments on top of an L2/3 transport layer
ii) Programmable Networking, or the ability to change and update packet forwarding and manipulation logic in real-time from outside the box (using something like OpenFlow)
iii) Cloud Management, or the ability both to automate network management and operations, and to compose network services as a component of a larger service across compute, storage, and networking
The combination of these three things will break apart the traditional vertically-integrated network stack, giving more choice to network operators and providing new platforms for innovation. It’s an exciting time to be in networking.”
SDNCentral: Can you summarize Brocade’s position on SDN? What customer segments are you targeting with SDN?
Keith: “We think the majority of early-adopters of SDNs will be those organizations who see it as a component of their larger Cloud Services strategy. Whether they are public cloud providers, carriers, the Web 2.0 companies, or large enterprises building out their private cloud architectures, it’s these organizations that have the most pressing need to solve some of these problems, and have the capabilities in-house to operationalize the new technologies needed to solve them.”
SDNCentral: Can you compare and contrast your SDN strategy versus those of other large SDN vendors? What’s unique about it?
Keith: “Brocade has been very active in the SDN community for years. We were one of the first major networking companies to publically endorse OpenFlow in 2010, we’ve been demonstrating our technology to customers since late 2010, and we recently were selected to be the networking supplier for Internet2’s new 100GE OpenFlow-based backbone.
The Internet2 win highlights one key area where Brocade has a different strategy than other vendors: the migration problem. How do I as an operator transition from my legacy network to a software-defined network? To do that, we think most initial production deployments of OpenFlow will use OpenFlow as an augmentation to traditional forwarding, not as a replacement for it. Brocade’s true Hybrid Mode support lets you maintain your traditional forwarding policies, and all the established operational processes to keep that traffic flowing, and then selectively enable OpenFlow on particular flows. Unlike some other vendors who force an “either/or” answer, Hybrid Mode lets organizations bring OpenFlow into their environment at their own pace.”
SDNCentral: Brocade already has an Ethernet Fabric. Why the interest in SDN at Brocade? What does SDN bring to Brocade customers that aren’t already in the portfolio?
Keith: “The move from traditional legacy architectures to fabric-based architectures has been in progress for a number of years now, and is now pretty broadly accepted as the right architecture for building out cloud data centers. At Brocade, we’ve now got over 700 production deployments of our VCS Ethernet Fabric technology. Ethernet Fabrics bring a much higher level of native automation to the network, increasing reliability and reducing the complexity and cost to operate your network. Ethernet Fabrics are also significantly more efficient in managing and directing traffic than traditional networks. Both in the lab and in production environments, we’ve seen people get a 30% efficiency increase in their network by moving to Ethernet Fabrics. That’s a significant increase in the amount of traffic your network can handle, leading to some pretty significant cost savings.
Coming back to SDN, the real question is what kind of network do I want to build my SDN on top of? To maximize the benefit of network virtualization (using, for example, an overlay networking technology like VXLAN), you want to be able to take advantage of the entire bandwidth of the network, not some fraction of that bandwidth dictated by inefficient legacy protocols. Similarly, automation within the network for things like discovery and switch configuration reduces latency and the potential for error. Ethernet Fabrics are the best foundation for an SDN.”
SDNCentral: How do you see SDN and /or OpenFlow manifesting itself throughout Brocade’s product line? What about your storage products?
Keith: “Brocade’s MLX line of routers supports OpenFlow at 100GE line-rate speeds with true Hybrid Mode today. Over the next year, we intend to extend that technology across the rest of our data center LAN portfolio. We’ve also made some recent announcements around VXLAN support in our portfolio, and again, you’ll see that extend across our data center portfolio over the next year or so. On the storage side, we’ve got some interesting ideas on how SDN technologies can be used in storage networking, so stay tuned on that front.”
SDNCentral: Brocade announced an interesting VXLAN Gateway capability for ADX? What customer problem are you solving? How is this unique?
Keith: “Overlay networking protocols like VXLAN and NVGRE are pretty interesting – they open up some new opportunities to architect more flexible and scalable data center networks. We recently announced VXLAN Gateway support on our ADX application delivery switching platform in support of our overall network virtualization strategy. Some of our biggest customers are embracing network virtualization using overlays, and they came to us asking us to solve two key challenges they saw. First, there needs to be an efficient, scalable, and cost-effective way of getting traffic from the un-virtualized Internet up into the overlay network (and vice-versa). Given its location in the network, and its hardware and software architecture, an ADC is the logical place to do that. Second, network services like ADCs and firewalls need to natively understand the overlay networks to be able to provide application policy management and enforcement. ADX is the first ADC to speak VXLAN, allowing it to natively participate in the overlay network.
Services natively speaking VXLAN are particularly interesting, as it is a more elegant solution to some traditionally challenging traffic management problems. The simplest deployment scenarios for an ADC are those where the ADC shares a subnet with the real servers. As networks have grown, we’ve moved the ADCs out of the server segments, and hung them off the aggregation layer. While a more scalable architecture, it also brings more complexity, forcing the use of things like Layer 3 DSR, Source NAT, etc. VXLAN simplifies out all these complexities, because with VXLAN, you can have a L2 adjacency with any server in the entire data center (or in a remote data center). This has the potential to be very powerful in a large data center.”
SDNCentral: Network virtualization via overlays is all the rage with VMware’s acquisition of Nicira. Is this the primary SDN use case you see or are there any other Brocade SDN customer use cases that you feel are particularly interesting or exciting? Can you share some details?
Keith: “Network virtualization in the data center using overlays certainly is one important use case for SDNs. There’s a lot more than that. We think Internet2’s WAN network virtualization use case using OpenFlow will be a model for a lot of carriers and research institutions. We also think there’s value in certain traditionally complex network problems, like complex packet replication and steering, and scalable network analytics. Every person I talk to on the business side of our customers recognizes that there’s an immense amount of valuable information contained within the network, but that information has historically been too costly to extract. Who’s using the network? When? Why? What’s their experience with the network? How can I recognize when that experience is being compromised, and remediate in real time? We think SDN is providing some new ways of looking at that problem, and when coupled with some of the things happening in the big data arena, create an opportunity to unlock that information and intelligence within the network.”
SDNCentral: How do you view the relationship between controller vendors and hardware vendors, like Brocade? Where do you see value created?
Keith: “It’s an exciting time in networking, because the unstoppable move to cloud architectures is creating so many opportunities for innovation. On the platform side, there’s a ton more work to do to build out the programmability and flexibility ultimately needed to achieve the SDN vision. Coming back to that migration problem again, there are also whole new sets of migration and troubleshooting technologies needed. On the software and controller side, there’s a ways to go to continue the standardization of interfaces that groups like ONF have been driving. In particular, we believe the northbound interfaces on the controllers need a level of standardization to match what OpenFlow is doing on the southbound side. It’s the applications that sit on top of the controller, far more than the controller itself, that create real value. For the open networking movement not to stall, standardization of those northbound interfaces is one of the next big issues for the community to tackle.”
SDNCentral: What do you think will the role of virtual switches in the datacenter? How does Brocade play in a virtual switch environment?
Keith: “Virtual switches will be an essential component of the virtual data center. Brocade has a long history of partnership with VMware, and our strategy is to continue to partner in the area of virtual switches. We’ve already integrated our technology with VMware’s to do things like auto-provision the fabric when a new VM comes online, dynamically reconfigure network forwarding when a VM is moved from one data center to another, and most recently exchange network topology and policy information to enable VXLAN support on our ADX.”
SDNCentral: What can customers expect from Brocade in the future around SDN and networking in general?
Keith: “At Brocade, we believe SDN will be a transformative change in how we architect, build, and operate networks. It will open up traditionally closed, single-vendor networks and bring competition and innovation to bear on some historically unsolvable networking problems. This will bring massive benefits to network operators, giving them back control of their network, and letting them use the network to innovate in their business, not be held back by the network. This also brings significant opportunity for us at Brocade. As such, you’ll continue to hear a lot more from Brocade over the coming months and years as this space evolve.”
SDNCentral: What do you tell a prospective customer who’s thinking about SDN what they should be planning to do now and in 2013 to implement SDN?
Keith: “First and foremost, ensure that your networking supplier has a publicly committed roadmap to SDN that includes support for open standards, and isn’t either developing a proprietary, closed approach, or relegating their SDN development to ‘experimental’ releases. You want to make sure that any equipment you buy today can get you to where you need to go. Second, identify one or two projects that solve a particularly painful architectural or operational challenge facing your organization. The objective is to get started, so the largest, most painful problem isn’t necessarily the one to bite off first. Finally, focus on the migration problem. Look for technologies like hybrid mode as a way to maintain an existing operations model for much of your traffic, and focus your SDN efforts on the specific applications that add value to your network today.”
SDNCentral: Thank you very much for your time, Keith!