SDNCentral is co-organizing (with Tech Field Day) a must-attend symposium on the software-defined data center on September 10, 2013. Be sure to join us in Santa Clara, Calif — click here for details. In the lead-up to the event, we reached out to some of the speakers and panelists to get their thoughts on the transition to the next generation of data center architecture.
For the first installment of this two-part interview series, we spoke with:
- Don Clark, director of business development, NEC
- Sunil Khandekar, CEO, Nuage Networks
- Howard Ting, VP marketing, Nutanix
Be on the lookout for Part 2, which might even include companies that don’t start with “N.”
What is the software-defined data center (SDDC), and what benefits does it promise?
Don Clark, NEC: The traditional data center design was really built around principles of isolation, but w/server virtualization and storage virtualization and multitenancy, there’s a strong need for increased agility that matches the capabilities and the flexibility of server and storage virtualization products. To us, the SDDC really has those aspects of flexibility while still maintaining that level of isolation.
Howard Ting, Nutanix: From an economic standpoint, we’ve reached a point where the data center network has to become more efficient, and it has to become more predictable. When you decouple the control plane from the data plane, it creates more flexible infra, and you’re able to better economically model what infrastructure you need.
The SDDC has the promise to greatly simplify IT as well. So there’s economic motivation, technological motivation, and an operational motivation.
Sunil Khandekar, Nuage: To us, it’s about choreographing the network so it becomes as consumable as the compute resources. What I see, based on my discussions with CIOs and IT admins of large enterprises, is that for the devops guys or the groups that want to deploy apps, the “ask” is very simple: “We want to deploy our application onto the network.” They don’t want to worry about what’s happening on the network. All they want is an infrastructure like AWS [Amazon Web Services], which is what everybody uses today.
What’s an example of an SDDC in use today?
Don Clark, NEC: NTT Communications’ Enterprise Cloud that they launched last year is a great example. NTT’s portable data center services are quickly provisioned and allow customers to go into a portal and manage their services directly — which would traditionally take a ticket and a couple of days’ turnaround time, at a minimum.
Howard Ting, Nutanix: Every one of the customers that Nutanix has, you could argue, has deployed an SDDC, because our solution is entirely software-defined. We use commodity hardware — commodity x86 servers with commodity SSDs and hard drives — and move the storage into the server so all I/O is localized. Our architecture is meant to eliminate the network as much as possible.
It’s not like these companies are deploying the full vision of the SDDC. They’re definitely experimenting. We’re talking the top of the Fortune 500 or global 2000s — and medium sized enterprises as well, and a couple hundred smallish enterprises.
Don Clark, NEC: In the traditional data center network, because we rely on traditional Layer 2 or Layer 3 protocols that are distributed, and they don’t have a common control plane, it really inhibits the agility necessary for an SDDC.
SDN can structuralize the control of the network and provide services dynamically while still maintaining the agility and reliability that’s required. It also exposes a single point of control that can be managed thru an API and help them through other network functions, so it’s very essential to the provisioning of NFV as these services evolve.
Howard Ting, Nutanix: We’re not as involved in NFV, but we do some SDN-like tricks. We do leverage the network for clustering and for creating a large pool of storage resources, but we commoditize it, so we use just standard 10G Ethernet to form that big pool of resources.
How do you see NFV and SDN relating to one another in the data center?
Sunil Khandekar, Nuage: NFV is a perfect use case for SDN, especially for the telco in cloud deployments. Telcos are looking to take advantage of the virtualized infrastructure, and this presents an opportunity for them to take a look at the network functions that have been delivered thus far, like IMS [IP multimedia subystem infrastructure] for example — to virtualize them and deploy them on x86-based platforms as they need them, as the capacity drives them to instantiate more, rather than build it with a certain capacity in mind.
A network that connects these requirements as they’re instantiated is critical. That’s where SDN comes in. So, that’s how the two come together.
What’s the role of software-defined storage (SDS) here? What does that term even mean?
Howard Ting, Nutanix: There’s been tremendous buzz and probably overmarketing around that term, but I think very few companies have actually developed and are selling a product right now. It’s not just about being “software-defined.” It’s about being converged … about having the ability to scale out.
We’ve seen on the networking side that there’s been more buzz and more media coverage. In storage, you’re going to see more deployments.
Sunil Khandekar, Nuage: There’s a clear movement going away from specialized adjunct networks like SANs to IP-based storage. Ultimately, it is about providing seamless access to the storage in a highly efficient way, to utilize software techniques that can partition network storage and make it available to applications as they need it. Again, the network, as you might imagine, plays a very big role in making sure that does happen.
I don’t think that means there is a need for everything to be vertically integrated. In fact, it’s the opposite. You have to make sure the network is able to adapt to a request without consideration as to what certain type of device is there; it could be several specialized appliances. It’s about the functions that are being provided.
What does the SDDC mean for the people involved — the network operators or development-and-operations (devops) people?
Don Clark, NEC: The employees within the data center will be able to add a lot more value to the operations instead of maintaining the current capabilities — a lot of the tasks in today’s data center are really driven by configuration and change management. The way I see the world changing is not unlike the way systems operators became virtualization operators in the last five years.
Howard Ting, Nutanix: It’s definitely a disruptive trend, but I think that’s just the nature of IT. There are constant disruption waves that happen every decade or so. We had no virtualization experts 10 years ago, and those people are the most high-paid, most-in-demand IT resources we have in companies now.
Obviously, for the people who don’t adapt and learn new skills, it absolutely has the potential to make their skills obsolete. But that’s just the nature of the industry that we all play in.
Sunil Khandekar, Nuage: For devops, it’s a dream come true. They want rapid deployment of apps when they need them; they don’t want to wait weeks or months. For the netops guys, it’s really changing focus from being tactical to being much more strategic, to not have to worry about mundane tasks that have to do with spreadsheet management. They want full visibility into the network and they want to provide their customers with supreme SLAs.
What should we be watching for at the SDDC?
Sunil Khandekar, Nuage: This will be the best of the best in terms of who understands the technology. We’re going to see some meaningful discussions in the conference. From our perspective, we know the work that has been done thus far aligns very well with this SDDC idea of automated compute and network infrastructure, so we are looking forward to having those conversations.
Don Clark, NEC: You’re going to see more and more use cases that are being deployed , leveraging the unique capabilities of the virtual data center.
We’re seeing three use cases of high interest for customers. First is the ability to deploy both test environments and production environments with the same characteristics — the same IP [address] ranges, the same VLAN markings — without having to redo everything as you move to production. The second use case we’re seeing is being able to apply quality of service for service delivery. And then a third use case is around this idea of service pooling — for example, being able to do things like we’re doing with Radware, rolling a DDoS appliance into an appliance pool and not having it be in-line.
Howard Ting, Nutanix: What the Symposium promises is to bring lots of different thought leaders together, to really have a kid of State of the Union about where we are with this big movement — and I call it a movement because it’s going to take many years before this is mainstream. It’s a great checkpoint on where we all are today, and we can get an objective view of what projects are really ready for prime time and which are science projects.