It's too easy to think about networks in terms of technology. To most people in the world, "the network" is whatever connects them with what they want. They don't think of switches or routers, hardware or software. They are buying experiences, and that's a fundamental change in networking. It might very well be what separates the "New IP" from the old, so let's look at IP from the outside in and derive what it must become from what users want it to do.
The basic difference between experience networking and traditional networking is that experiences are hosted somewhere. A network that has to support communications has to support universal, relatively uniform, connectivity. An experience network connects you to your experiences, which are probably not far away.
Any content element popular enough to be profitable can be cached in every metro area where it's likely to be consumed. We see that now with content delivery networks (CDNs). It's also very likely that as cloud computing gets more pervasive, more hosting is going to spring up in each metro area and user-to-cloud connections will become shorter.
Changing the nature of the network edge
Connecting to hosted experiences changes the whole nature of the network edge. Instead of worrying about which of billions of Internet endpoints a user might want to talk with, you simply deliver content from a nearby metro cache or make a fairly static connection to a cloud data center. A given user might have a dozen content relationships in an hour.
The impact of this on software-defined networking (SDN) could be enormous. SDN and OpenFlow allow a central controller to set up forwarding rules as traffic is presented. Obviously that wouldn't be workable if we assumed a user addressed the whole of the Internet, but could it support those dozen content relationships? I think it could.
This doesn’t mean that our future user loses the Internet, only that we create what are effectively two parallel pathways to getting what we want -- one for high-volume, high-value content and one for everything else. We do this already at the CDN level; I'm just suggesting we could extend that principle all the way to the network edge.
And that could change everything, over time. Studies by Cisco Systems Inc. (Nasdaq: CSCO) and others show that nearly all the growth in traffic in the future will come from growth in experience delivery. We're calling less, doing less with "connections" as we do more with social networking, content delivery and the cloud. All of those are hosted and all could be delivered using SDN technology right to the user.
Moving to a mobile future
But this isn't the end of our experience-driven changes. Mobile is the way of the future, right? Mobile services are almost uniquely "contextual" in nature, meaning that mobile search and mobile applications in general need to be aware of where the user is and what they're doing. We're adding the notion of personal digital assistance to this mix, things like Apple's Siri or Microsoft's Cortana. These assistants will change the flow of information even for non-content applications.
We expect to be able to ask these assistants questions in the future, and get answers. To make this work, mobile users will interact with an agent process in the cloud that will "ask" information and content resources for things and then pass them on to the user. Think about that for a moment. If the agent in the cloud talks to information resources in the cloud then where are the information flows? In the cloud. The user has one connection to the digital assistant and no other direct connections at all. Networking is now intra-cloud and into-cloud. The whole of the Internet starts to look more like a data center than a network.
Now we can introduce NFV concepts. If a user's needs are fulfilled by ad hoc collections of information and functionality, what better way to deploy that than NFV? We can spin up everything from our digital assistant/agent to information-culling processes on demand. We can host them close to the user, close to the content, close to the processing resources, whatever makes sense. Because everything is totally automated, the costs of personalization are reasonable. New service revenues are generated not by pushing new or different bits or pushing them in different ways, but by creating things that are valuable because they connect with what everyone wants, what everyone is doing.
Scaling networks functionally
I think this plays well into SDN guru Scott Shenker's "SDN2" vision. You have a network edge that is not only more software-defined, it's more software-hosted. If that network edge serves both as a hosting point for intelligence and as a conduit for specialized traffic, it is best deployed as a computer with network software. The knowledge of different information resources, different agent processes, is confined to the edges of the network because that's where the requests by users are separated. Once Agent A decides it needs to talk to Agent B, we're on aggregate pipes between them.
Looking at things from the experience side demonstrates that we have to scale networks functionally if we want to scale their value. It also demonstrates that functional scaling will increasingly make "the network" into "the cloud", and finally into the data center.
— Tom Nolle, President/Founder/Principal Analyst, CIMI Corp.