Intel takes both sides in the 4G debate – Cloud-RAN or edge cloud?
The biggest change in 3G+ and 4G networks, compared to their predecessors, is the amount of data they have to handle
Published: 25 May, 2011
The biggest change in 3G+ and 4G networks, compared to their predecessors, is the amount of data they have to handle. That is not just about processing consumer data traffic, but supporting a huge increase in the intelligence of the network itself – how it identifies different classes of data or subscribers, for instance, and matches them with appropriate services, tariffs and speeds (or offloads them). To deliver a strong mobile quality of experience, carriers need to handle rising tides of signalling, web and apps traffic, as well as prioritizing certain transactions and data types. Most agree that this challenge will be met by designing new networks around a far denser mesh of small cells, bringing the signal a great deal closer to the user. But while the antenna and radio may get nearer to the device, where does all the vast amount of data processing and network tasks go on? Does that, too, need to be pushed as close to the end user as possible? Or at the other extreme, should it all be centralized in huge core network servers, and even virtualized in the cloud?
The answer depends on the particular business model, spectrum position and population density of the carrier, and most will use a combination of techniques in different locations or services. It also depends on which vendors’ voices get heard most loudly by those carriers, and whereabouts in the network their key strengths lie. Some are obvious – we can expect talk of centralization and virtualization from Cisco and enterprise integrators like IBM; while RAN suppliers whose core network offerings have been under pressure may prefer to push processing closer to the network edge, as Nokia Siemens has done with its gateway strategy, as well as specialists like Stoke.
Other suppliers have more complicated agendas. Alcatel-Lucent is gaining ground in central and edge routers, and is a huge proponent of LTE networks built around tiny cells with at least some intelligence in the base station. Yet ALU also made one of the decade’s most disruptive pronouncements so far, when it called the effective death of the base station with its lightRadio launch. The ultimate goal of this is to leave highly integrated, highly commoditized antenna/radio units on the towers or roofs, and put all the intelligence and baseband processing into central cloud servers supporting thousands of cell sites.
Another important player with a foot in both camps is Intel, whose general manager of the Communications Infrastructure Division, Rose Schooler, discussed two different developments this week, representing both extremes of the edge versus central cloud debate. One is the chip giant’s project with China Mobile and an unnamed OEM (probably ZTE) to develop a Cloud-RAN platform for the operator’s TD-LTE roll-out, virtualizing the baseband processing for thousands of deconstructed base stations. The other is a collaboration with UK femtocell start-up Ubiquisys, to put an Atom processor into a small base station, bringing not just network tasks but significant data processing and storage to the very edge of the RAN.