The Flat World Theory

As far as core networks go, Clearwire's is pretty simple. Its base stations link to access service network, or ASN, gateways, from which IP traffic disappears into the Internet cloud. The long-term evolution networks planned by AT&T Mobility, Verizon Wireless and other global wireless operators have a similar architecture: Traffic from the base station (or the evolved Node B, as it is now called) makes a single hop to a service architecture evolution, or SAE, gateway before hitting its first IP router.

While the telecom industry certainly can't resist creating new acronyms, there's a noticeable lessening of the number of acronyms describing nodes in the next-generation mobile network. Gone are the base station controllers (BSCs), the radio network controllers (RNCs), the serving GPRS support nodes (SGSNs), gateway GPRS support nodes (GGSNs) and packet data serving nodes (PDSNs).

Instead, the next generation of mobile networks is embracing the notion of a flat network architecture, one that runs entirely via IP and collapses the complexity of a mobile network into the base station and a single core gateway. The wireless industry is essentially modeling the new mobile network off of the IP architectures that have driven data networking for two decades — the assumption being that such designs will allow mobile operators to deliver more capacity, more efficiently, to their customers. But what is that assumption based on? Collapsing the network core doesn't fundamentally increase the amount of data that can be shoved into wireless air links. What good will excising a few boxes from the network achieve? Why is a flat IP architecture better?

To understand why is to understand how mobile networks have been designed up to this point — for circuit-switched voice. Like their wireline complements, wireless networks were designed in a hierarchal fashion to aggregate, authenticate, manage and direct calls. A BSC aggregates calls from multiple base stations, allocates radio channels, enables handoffs between base stations and passes on calls to an even more centralized mobile switching center. As packet data networks emerged, they were overlaid on the existing voice-centric architecture, using the BSC for the same mobility management functions and adding the SGSN and GGSN in the case of GSM/UMTS and a PDSN in the case of CDMA to route and manage data sessions, as well as to connect to the Internet or appropriate IP network.

The system functioned just fine when packet data was a small amount of traffic on the network, but as 3G radio technologies improve and more powerful smartphones and feature phones flood the market, operators are seeing data usage from their customers skyrocket. Data traffic already may have outpaced voice traffic on most operators' networks, but data revenues haven't kept up. Operators are faced with a quandary: They have to expand their network capacity to meet the primary driver of growth in their networks — data — but the return on those investments is diminishing.

A flat network architecture removes that voice-centric hierarchy from the network. Instead of overlaying a packet data core on the voice network, a separate and much-simplified data architecture can be implemented that removes the multiple elements from the network chain. BSC functions are collapsed into the base station itself, which communicates directly with what is essentially an IP router or gateway over a carrier Ethernet link.

The capex savings and the operational efficiencies of such a simplified architecture can be dramatic, said Maniam Palanivelu, senior manager of LTE marketing for Nortel Networks. “You have fewer hops and fewer nodes on the network,” Palanivelu said. “Overall the fewer nodes you have, the less infrastructure you have to deploy, the fewer elements you have to connect. That's substantial capital and operational savings.”