Mobile data services: Core vs. edge

It's about architecture, and it's happened before. When steam engines were new, naval architects modified their designs to accommodate the use of steam engines in ships that were still basically sailing ships. As long as deployments remained limited, this jury-rigged combination worked just fine. But when it became clear steam engine power was required on a widespread basis, naval architects redesigned their ships with new hull shapes, new materials, and new methods to make the most of steam engines, and ultimately stopped designing ships with sails.

The mobile Internet is at a similar stage, and it's still about architecture. It isn't enough to make the Internet mobile, any more than it was sufficient to put engines in sailing ships. We need to reshape the Internet infrastructure to match its mobility.

Fixed-network solutions for mobile-network problems have gone about as far as they can go. What worked in the fixed world isn't scalable enough to meet the expected demand for mobile data services, nor is it flexible enough to deliver dynamic, localized and personalized services.

A brief history of edge and core

Until recently, the vast majority of data products were designed for and deployed in fixed networks. Routers--fairly dumb devices by current standards--became the workhorses of the Internet in the late 1980s and early 1990s. They quickly became more powerful, and by the end of the 1990s, routers began adding the smarts required to move to the edge of the fixed network.

In a fixed data network, anything requiring intelligence happens at the edge. Edge devices aggregate and shape traffic, terminate DSL and dial links, and manage the gritty service details of a relatively small group of subscribers. Putting intelligence on the edge of a fixed data network clearly is the most efficient solution because in the fixed network the edge is where service policies are known. Customer services, restrictions, permissions and billing options have to be "remembered" somewhere; so putting this service information out to the edge of the network, closer to the customer, is the logical option. And if a customer in Orlando receives data from Miami, it's better to forward his traffic directly rather than routing it through Chicago.

The superiority of a smart-on-the-edge, dumb-at-the-core network architecture for fixed data networks was so clear to so many people that an entire sub-industry grew up in the late 1990s, devoted to building the devices for the network edge. The very word "edge" acquired a certain cachet. In the boom-driven word association of the time, the minds of vendors, venture capitalists, and network architects began to link "edge" to "intelligence," "smarts," "leading" and more generally to value creation. To be on the edge was to be new, innovative, far-sighted and adventurous. And so it still is, if what one designs, builds or runs is a fixed data network.

Many are applying this fixed network approach to mobile networks by re-packaging fixed-data solutions for mobile data networks. Network architects, provided with virtually no optimized alternatives, are working hard to design, implement, and run their networks with intelligence built on products optimized for the fixed network edge. This solution works well enough as long there aren't too many subscribers and the expectations of those subscribers remain mercifully low.  However, it won't function with large numbers of subscribers, and may limit the opportunity for mobile operators to deliver compelling services and billing models that could help drive uptake.

Mobile service core requirements

While the users of mobile data networks may actually be the same as those on fixed data networks, the subscribers are not.

Are the users and subscribers the same people? Well, in many cases, yes, they are. But on a mobile data network, users are often subscribers, as well. A remote sales office with eight people will typically account for one subscriber of a fixed data network, but the same office will yield eight individual subscribers for a mobile data network. That hypothetical sales office may have eight sets of mobile services, permissions and billing options.

Similarly, a corporate headquarters that might be seen as a single subscriber in a fixed data network may account for hundreds, or even thousands, of individual subscribers in a mobile data network.

Not only are mobile subscribers more numerous than fixed subscribers, they behave differently. They don't browse much. Instead, they are on the go, and they need very specific information, right now.

Consider an Italian executive disembarking at the airport in Atlanta after a long flight from Rome. If he decides to read The New York Times online from his study in Rome, he will not expect to read it in Italian. In Atlanta, on a mobile phone, he will not be in an experimental mood. He will want to see his customary portal, and receive his information in Italian. He will want to proactively access his applications, get his messages, check his e-mail, confirm his appointments and return flight, and get on with his life.

A mobile network provider will need to dynamically link this subscriber to his services, route his messages to him and charge him appropriately for the value of the services he receives. This mobile operator will be the subscriber's vendor--in this case, back in Italy (far from the radios at the edge of the mobile network in Atlanta). Therefore, intelligence can't be too close to the subscriber. There are too many subscribers, and they're moving all over the place (including into networks owned and administered by other operators).

Complicating the issue is that these existing edge devices, smart as they are, soon won't be smart enough. Vendors are modifying routers and DSL subscriber management platforms for initial 2.5G network buildouts, meeting a minimum mandatory set of features--wireless session termination, mobility management, and rudimentary packet-routing and transfer functions. But the architectural paradigm of fixed data networks, in which services are applied to a link and subscribers are charged for that link, doesn't apply to mobile. What's missing in all this retrofitting is the ability to link a subscriber's identity with a set of personalized data and charging services that allow the network operator to sell him new services that ultimately will drive mobile data revenues.

The subscriber-aware mobile service core

A subscriber-aware mobile data network would rest on a layered architecture, which would rely on Internet Protocol (IP)-based core networks built with the best available packet transport technologies. Packet transport--routing, switching, time-division multiplexing, and optical--constitutes the bottom layer of the architecture, and this layer is common with that deployed in today's fixed IP networks.

The subscriber-aware services layer is the next one up, where the network links its subscribers to a highly stateful array of information regarding their services, permissions and billing options. In this layer, the network processes traffic according to each subscriber's billing, addressing, security and forwarding policies. The subscriber services layer (often implemented in mobile service delivery systems) interfaces to an array of mobile middleware including subscriber databases, presence databases and mediation platforms.

This mobile subscriber service delivery function is vastly different from routing, where the basic function is to examine the IP destinations of packets and forward the packets to next hops intended to move the packet closer to the destination.  Unlike mobile service delivery, routing is not at all stateful, and does not require knowledge of a specific subscriber service policy.

Core intelligence for personalizing mobile data services

For example, our visiting Italian executive, stepping off his plane in Atlanta, would carry with him the service profile and billing information residing with his own network provider. At the core of his provider's network would be the knowledge that this subscriber is postpaid, has secure mobile enterprise connectivity, should get free access to specific content-provider-sponsored sites and should get pushed information of interest (sales updates from his enterprise, or a continually updated Alitalia flight schedule between Rome and cities he visits most often). His mobile terminal is company issued, and so his network provider knows that he is not allowed to access costly premium sites. No matter where in the world he is, our executive can access certain applications in his corporate network, but not others. Likewise, his network's core knows all about the security measures that protect the executive and his company's networks from unauthorized access and interference. Finally, because our Italian subscriber is roaming, it is important to capture this fact (along with the roaming partner) as part of the charging data, as all charges will be based on higher roaming tariffs.

To support these functions, the network must be able to detect where our subscriber is, whether he is connected (present) or not, associate him with his personalized service policies and dynamically determine the context of his session (his access network identity and capabilities, the services he is accessing and the content he using). And in the case of a subscriber using prepaid data services (already common in many regions of the world and increasingly common within the United States), a tight integration is required between the mobile service delivery environment and the real-time charging systems that authenticate the user and debit his prepaid account. 

A handful of vendors, including Megisto, have designed their infrastructures for subscribers and services rather than links, with the understanding that the brains of a mobile data network should be at its core, and with the processing power required to effectively deliver personalized services. Such infrastructures will offer operators a flexible platform from which to customize new services for discrete groups of subscribers, or for single subscribers. The new core infrastructures make maximum use of a network provider's unique knowledge of its customers as individuals--who they are, where they are, what services they use, how they are charged and whether they are connected or not--to craft these new services while lowering capital and operating costs.

Subscriber state awareness required

Key to effectively delivering these services is a design that can manage vast amounts of highly personalized subscriber state data. The services provided to a subscriber may vary depending on whether the subscriber is coming from a 2.5G network, a 3G network, a public WLAN hot spot, or whether the subscriber is at home or roaming in a third-party (and more costly) network. The services provided may also be varied based on the state of the subscriber's account. A prepaid subscriber may be redirected to a business support system portal that encourages them to "top off" when out of funds and perhaps given access to the operator's walled garden to encourage them to remain on the network--but only if accessing the network at home where the operator does not incur roaming charges for this subscriber. Such services are highly complex and stateful--and require effective access to vast stores of memory. And the state is often critical accounting data so any design must support resiliency so that this data is not lost due to a network or component outage.

The mobile network operator wishing to offer new services over a network based on retrofitted fixed networking gear must be sure those services work and are in demand, because expensive devices dedicated to specific value-added services will have to be added to their core networks before the services can be delivered. There will be no room for experimentation and little room for error. Now, if the network's brains were at its core, based on equipment designed to deliver mobile services, these new services could be flexibly offered to those most likely to want them and quietly ended if they didn't work. Such a network would be focused on its subscribers, not its links. Such a network would be subscriber aware.

Core beats edge for ROI

Every advocate for a new technical solution talks about saving capital and operating costs, but here the case is easy to make. A core solution for the mobile data network involves fewer, less expensive devices. These devices, because they are tied to people and not places, and because they are service aware, can be used to design, test, deliver and alter services without having to conduct a detailed and lengthy analysis for each new service.

By placing intelligence at the core of the network, operators can offer customizable mobile data services without putting their companies at risk in return for a small gain in revenues.

Carol Politi is Co-Founder and Vice President of Marketing for Megisto Systems.

Visit Megisto Systems online.