Mobile services are the service provider networking sweet spot. For years Wall Street has said that capex for mobile would be up while wireline would be flat-to-down. Given that it's harder to adopt new technology where you're curtailing spending, it's logical to assume that the New IP will emerge in mobile. But what will this "New Mobile" look like?
Handling the traffic
The actual traffic handling in a traditional mobile network is largely the responsibility of the Evolved Packet Core (EPC). What happens, in simple terms, is that an EPC element "anchors" a tunnel that represents a device at a point near the IP/Internet access gateway. The other end of this tunnel is moved around to point to the cell where the user can currently be found. If the user moves, the user end of the tunnel moves too, and the fact that the user moved to another cell is handled by the IP Multimedia Subsystem (IMS) part. The anchored end of the tunnel is what the Internet or other services see as the user's IP address, and since that doesn't change even as the user moves, EPC accommodates mobility without requiring applications support changing IP addresses in the middle of a session.
All of this was very logical 15-plus years ago when the concept emerged, but software-defined networking (SDN) surely offers an alternative approach. In SDN, a central controller sets up forwarding rules in each network device -- in this case, each device within a metro network that supports a bunch of cell sites. If a user moves from cell site to cell site, the IMS logic that registers devices would see that move, and simply command the SDN controller to reroute the user's traffic to the correct site by calculating the changes in route and sending forwarding-table changes as needed. No more tunnels and none of the complicated EPC elements.
Besides simplifying mobility management, the SDN model could also help with offload and content delivery. An SDN controller could "learn" from a DNS server or CDN network that the user wants a piece of content that's located in a cache somewhere, and create a route for that content to be delivered independent of where the rest of the traffic went. A video call could be routed for priority-level QoS, separate from the default path that web traffic or even voice might take.
Facing the challenges
You'd think this is the sort of thing we'd hear about at Mobile World Congress, but instead it seems like the focus is on the relatively simple task of converting the elements of IMS and EPC infrastructure from dedicated appliances to hosted form. In fact, most implementations of virtual mobile infrastructure have all the same elements, the same steps and the same dependency on legacy IP features that we defined 15 years ago.
I think this illustrates one of the challenges we face on the road to the New IP. We have to decide whether a New IP is "new" in the sense that we build services in a way that's unfettered by the limitations of physical Ethernet or IP devices, or whether we just build services using the same protocols and techniques but with hosted software instead of devices.
There is a value to using hosted elements for IMS and EPC, to be sure. We can make networks cheaper, and perhaps more resilient to failure or elastic in response to changes in load. But is that enough to justify saying it's "new"? If you put a traffic analyzer into such a network, we'd see nothing we don't see today, which doesn't seem all that new to me.
Getting to the 'new' IP
IP is a network protocol, and so it's hard, I think, to create a "New IP" without changing how network protocols work. That change doesn't have to affect the services or the user, but it should affect the way traffic moves and add value to services. Some of that can be accomplished by exploiting network functions virtualization (NFV) at layers higher than the network layer, where IP lives, but I think that SDN and OpenFlow are the keys to getting to a really "new" IP.
Virtual routing and switching are the critical steps in exploiting the potential of SDN, because they can add SDN-based route control to traditional IP or Ethernet networks. That lets us create a real "virtual IMS and EPC," without forcing services that don't yet justify new IP or Ethernet handling to be changed. These are the bridges from traditional IP to the new era, where software-hosted-and-controlled traffic handling and features will dominate Layer 2 and higher.
As we move from 4G to 5G and face increased spectrum pressure, we'll be using smaller and smaller cells, and more wireless backhaul and EPC technology. Standards that were developed in the previous decade, before mobile broadband and smartphones, aren't likely our best choice. Today, we have a rare opportunity, as technologies change and capex continues its shift to mobile, to merge metro optics with Ethernet backhaul and SDN-based "New IP" to create a better approach.
— Tom Nolle, President/Founder/Principal Analyst, CIMI Corp., special to The New IP