Jeff Doyle speaking at TFI2014
We’re happy to announce that one of our CO ISOC board members, and TFI2014 steering committee members, will also be joining us as a speaker for The Future of the Internet 2014: Defining Software Defined Networks: Jeff Doyle!
Jeff Doyle, Principal Architect at FishNet Security
Specializing in routing and MPLS, data center architectures, SDN, and IPv6, Jeff Doyle has designed or assisted in the design of large-scale IP service provider networks in 26 countries over 6 continents. He worked with early IPv6 adopters in Japan, China, and South Korea, and now advises service providers, government agencies, military contractors, equipment manufacturers, and large enterprises on network design best practices.
Jeff is the author of CCIE Professional Development: Routing TCP/IP, Volumes I and II; OSPF and IS-IS: Choosing an IGP for Large-Scale Networks; and is an editor and contributing author of Juniper Networks Routers: The Complete Reference. He also writes blogs for both Network World and for Network Computing. Jeff is one of the founders of the Rocky Mountain IPv6 Task Force and is an IPv6 Forum Fellow.
Take it away Jeff!
How Does the Network of the Future Differ from Today’s Networks?
You don’t really need to look into a crystal ball to see what tomorrow’s networks will look like. You really only have to look at today’s networking pain points. There are plenty of interesting new technologies arising every day, but the ones that will succeed are the ones that relive some of the “owies”:
- Human Error. A person touching individual network elements remains the primary cause of network problems. The more a complex system can be automated, without humans inserted between elements or between layers, the more predictable and reliable the system becomes. And speaking of reliability…
- Unreliability. This is where vendors start talking about five-nines and MTBF. But reliability has to apply to a system as whole, not just individual components. A network exists to support applications, so we have to take an application-centric approach to building networks.
- Unavailability. Rather than looking at how quickly a network can recover from a failure, we have to look at available our applications are over the network. Another way of looking at this is application continuity.
- Inflexibility. Yes, agility is a popular marketing buzzword. But it matters. Businesses cannot wait months, weeks, or even days for a network to be adapted to new applications. The network of the future must adapt almost instantaneously to changing application demands.
It’s surprising that anyone is still arguing about this one. IPv4 is done. It’s depleted. And trying to keep our networks limping along on this exhausted resource is increasingly like squeezing blood from a turnip.
It’s time to give up our NATs and our highly segmented networks. The network of the future will not be a dual stacked IPv4/IPv6 network. It will be an IPv6 only network. Complex address design and management, “security through obscurity,” address dependencies in applications, private addressing, address overlap, and VLSM will be as obsolete as IPv4.
Tomorrow’s networks will use one-size-fits-all subnets. They will have true end-to-end security practices. They will be easier to troubleshoot from an address perspective (yes, IPv6 is easier to work with than IPv4). They will have better mobility. They will have better multicast. Routing tables will be more manageable.
IPv6 makes all this possible. We move from the constraints of a highly depleted resource to a practically unlimited resource.
We are already seeing the benefits of automation in the data center, where the compute and storage elements are abstracted and controlled by an orchestration layer. Virtual machines and virtual storage are agile, mobile, and adaptable.
But the network is currently a roadblock. Operators still have to manually reconfigure the network to fit the changing needs of the compute and storage elements in a data center. This means high operational expenses, high rates of configuration errors, inconsistent configurations, and heavy delays.
SDN and NFV move us closer to an automated network, and while early efforts are mostly centered on data center environments we can expect the lessons learned to quickly move into enterprise and service provider networks. Networks will become abstractions, controlled by the same orchestration that controls storage and compute. Network elements can be deployed or withdrawn on demand as can stateful services such firewalls, load balancers, and cache engines.
Integrating the virtualized network into the same orchestration layer as compute and storage gives rise to network programmability. This is more than the scripting we often use to manage todays networks. The individual elements of the networks of the future – switches, routers, firewalls, load balancers, and so on – will not be individually configured using diverse operating systems and vendor-defined configuration statements. Instead, the network will be holistically configured with a programmatic “If… Then… Else” syntax.
Programmability puts a layer of abstraction between the humans defining network behavior and the network elements themselves. Configuration error is reduced, as are operational and training costs.
Application Defined Networks?
All of these trends help bring the network more readily in service of applications. Looking a bit further into the future, one can see a time when the orchestration layer controlling the network becomes an arbitration layer between the network and the applications. The orchestration layer knows, moment by moment, what network resources are available and what immediate demands the applications have. The network then adapts dynamically, and in real time, to changing application requirements: Application defined networks.
Humans then are removed entirely from the vertical stack. Our role becomes one of defining policies to the network, rather than acting as intermediaries between the applications and the network.
These trends are well underway, so the network of the future is not all that hard to predict. It will be more deterministic, more reliable, more available, and far more agile than the networks of today.
Make sure you Register Now so that you can discuss all of this and more with Jeff Doyle Friday, 22 August at TFI2014!