Loading the player...

Additional Videos

Ciena OPn Network Architecture
Length: 2:39

Packet Networking
Length: 3:07

Data Center Without Walls
Length: 2:58

WaveLogic 3 - Rick Dodd

view all videos

The Future of Networking (Length: 7:03)

The network of the future is a dynamic, programmable, agile, and adaptable common infrastructure for running applications and connecting places, people, and machines. All this connection inevitably increases bandwidth demand, but this network bends the cost curve down, decreasing the cost of networking while increasing bandwidth. Ciena's OPⁿ network architecture fits these attributes—optical packet, open, and programmable, to great scale.

+ view full transcript

STEVE ALEXANDER: Hey, welcome to Ciena Chalk Talk. I'm Steve Alexander, and we're here to talk about the future of networking. So networks are built to connect users to providers of products or providers of services. And in the past, we would generally build a network that was specific for a certain type of product or service. We would've built copper networks if we were going to do voice. We would've been coax networks for video. We'd have wireless networks. And we'd even have networks for distribution of products-- movies, music, news, magazines, and such.

All those different types of networks were built off very specific technology platforms. And all of these networks are changing. They're being replaced now, by one common network that we really think of more as a platform, an infrastructure platform, for future services and products.

OK, so now let's double click on that network infrastructure, that platform, and see what's inside. What you see is the combinations of server technology, storage technology, and what we'd classically think of as the network-- it's the connections amongst all these things. And what you're really talking about in terms of this platform is something that can run applications because you don't want to lose the attributes that you had when you had service-specific networks. You want to go to bring those forward into this new world, but you want them to be able to run on one common infrastructure.

So what would be an application that we might want to run on here? Here's a very simple one. Here's just network partitioning. I might want to partition my network for some simple administrative purposes, or better yet, maybe, I'm doing something like I'm covering, let's say, the World Cup.

And I want to isolate a portion of my network. And I want to make sure that it runs separately. I want to have it be very highly available. I want to be replicating everything inside of it because the last thing I ever want to do is drop that connection, drop those connections, and drop the World Cup service. And so, logically, I can partition my network.

So another application I might want to run would be a private line. Let's say I want to create, for this example, a very high capacity Ethernet service for somebody between two locations. Again, I want to run an application on top of this platform that then creates the service for my customer.

Another application I can run would be something that would be used, let's say, for data centers. Something that would be synchronization. Something that might follow the sun. So during the daytime, I'm setting up connections, for example, maybe 100 gigabyte Ethernet connection. They're connecting the data centers together. And in the background, there's a fiber channel that's running also.

At night, I may want the characteristics of that to change. I may want all that capacity to show up on the fiber channel link. I still want Ethernet connectivity, but it doesn't need to be nearly as big. Again, that's an application that runs on this platform that creates the service that ultimately my customers are interested in.

OK, so let's talk about how we build such a platform, an infrastructure platform. Because you want that infrastructure platform to be dynamic. You want it to be programmable. You want it to be agile. Again, you want it to adapt to the service and user experience you're trying to create. And you have to do this in an environment where we have incredibly accelerating demand for capacity.

You know, when we first started to build networks, it was really all around connecting locations together. Then we started to connect people together. That's the wireless networks. Now we're also building networks that connect machines together. And it's really that combination of locations, and people, and machines that's driving this tremendous demand for bandwidth.

Now we have to be able to build these networks in a cost-effective way. You can't have the network get 10 or 100 times bigger and cost 10 or 100 times as much. We have to be able to bend those cost curve down and let the network costs grow much, much slower than the network capacity.

So a couple of very simple principles you want to use in terms of minimizing cost. What we've captured on here is what you would call the cost pyramid. What you see, as you grow higher layers in the network, things cost more. It's a very efficient use of network resources to stay optical as long as possible-- much lower cost, much lower power.

But you need all these different layers to build a complete network infrastructure. You have to have all this to build this platform. But you really want to optimize how you use each one of these layers. And in fact, a couple of very simple principles.

The first one is you want to minimize IP touch. That's absolutely critical because IP touch is the most expensive thing you can do in network. The second thing you've got to do is, because you're going to be carrying tremendous amount of packet traffic, is you want to optimize around packet transport. It's those two key features is really going to help you in terms of optimizing the cost of the network.

And then there's four things that are really critical in terms of actually how you build the hardware that goes into the network and how you automate it so it becomes dynamic, so it becomes programmable, so it becomes agile. First thing, right down at the physical layer, you want to use coherent optics, digital signal processing. That's how you're going to solve all the propagation. That makes the signaling on the network independent of the type of fiber you've got in the networking, the age of that fiber.

Second thing you've got to do. You're going to use a distributed control plane for intelligence. That's how all the different network elements are going to communicate. It's how they're going to determine who's connected to who and what the topology of the network is.

The next thing you're going to do is use virtualized switching because you want to be able to encapsulate traffic in any way it needs to be carried in the network-- whether you want to do it with OTM, kind of like a containerized freight model, whether you're doing MPLS, whether you're doing Ethernet-- all of those different forms of encapsulation will exist in your network, but you want to do that with virtual switches. That makes a network as flexible as possible.

The last thing you need are APIs, the Application Programming Interfaces. That's how you're going to communicate with this platform. And within the platform, that's how the elements will communicate with each other.

If you do those four things, you also minimize IP touch. And you optimize around packet transport. Then you've bent that cost curve down. And your network can grow to be 10 or 100 times bigger. And you don't have to suffer the inherent costs associated with IP. So these principles that we've talked about, they're all contained in an our architecture called OPn-- OP raised to the n. Optical packet to great scale, open and programmable also to great scale. And OPn really does change the way you compete.