Examining Intel's conception of the radio access network
Intel aims to be everywhere in the radio access network, centralizing the baseband in the cloud and pushing applications out to the cell
Much attention is focused on Intel’s efforts to get its processors into the guts of the smartphone and mobile computing device, but Intel’s wireless ambitions don’t just lie at the terminal end of the mobile network. Lately Intel has been pushing its silicon into the radio access network (RAN) itself.
Intel is betting that the processor architecture that revolutionized the PC industry can do the same for the telecom industry, allowing wireless equipment vendors and operators to tap into the pace of innovation and the economies of scale of the computing world. Intel has joined several traditional telecom vendors working with China Mobile to develop a cloud-RAN architecture that would remove the base station from the cell and distribute its functions within a carrier cloud, setting its X86 architecture for a new task: baseband processing (CP: Intel’s cloud expands to cover the wireless network).
The China Mobile project is a long-term research trial, but Intel has more immediate intentions for the RAN. While it is pushing baseband processing from the cell core, it aims to push application processing in the opposite direction. It is working with Ubiquisys to embed its PC processor line into the vendor’s small cell base stations. Base stations are no strangers to applications processors, but Intel and Ubiquisys plan to take it to the next level, doing away with the integrated baseband-applications chips that feature in most pico and micro cells to create a dedicated applications engine that can perform higher-order computing at the network’s furthest perimeter. “There is more and more demand to put more intelligence on the network edge,” said Caroline Chan, wireless segment manager for Intel’s embedded communications group.
That statement may seem contradictory, considering Intel is alternately working to push baseband intelligence away from the network edge, but Chan explained that the needs of the network change depending on the topology used. With the small cells developed by Ubiquisys--basically as a step up from a public-access femtocell--backhaul resources will always be limited. Rather than link every small cell with fiber, operators will be looking for copper or even point-to-point wireless solutions, which could lead to a bottleneck in the network. Meanwhile macro-cells are all moving to fiber Ethernet backhaul giving them the capacity necessary to move their computing and baseband power further into the network.
Ubiquisys believes it can use Intel chips to effectively build a content delivery network into every small cell. Popular content can be cached at the cell site, and that content can vary depending on where the cell is located, said Will Franks, chief technology officer and co-founder of Ubiquisys. At a soccer stadium, the rare goal scored by one of the teams is an often downloaded and streamed video even on the stadium premises, Franks said. A cell with advanced processing capabilities could round up all of the related videos from such a score and cache them at site, where they can be streamed to attendees without overtaxing the backhaul network. The same principle could be applied to geo-local content. Google Maps street view, for instance could save its hi-rez photographs of local geography within the cell, drawing on its cache whenever a nearby street view is requested by a phone within the cell, Franks said.
Application processing could be used to transcode and optimize video at the cell site downgrading or upgrading stream quality depending on congestion. Operators could implement policy controls on a cell-by-cell basis. It can also be used to optimize upstream traffic. Photo uploads or other non-real-time content, for instance, could be cached at the cell to make way for real-time services such as video chat and later uploaded into the network, Franks said.
“With any of these types of [heterogeneous] networks, you need to run applications where it makes most sense to run them,” Franks said.
There has always been application processing at the base station. Even the lowliest femtocell needs a CPU to run its non-baseband functions. But those solutions have tended to be integrated chipsets that designate a certain modicum of resources to applications processing, but nowhere near the power needed to run the advanced services Ubiquisys and Intel are exploring, Franks said.
The same trend is happening in the device space. For most of their history, phones ran on integrated baseband-applications processors, but with the advent of the smartphone and mobile computer, devices are now shipping with chipsets containing dedicated applications processors. Ubiquisys will take the same approach with the small cell, paring Texas Instruments baseband processors with Intel CPUs ranging from the Atom to the Xeon class.
That huge difference in computing power isn’t determined by the size of the cell, but rather by the number and might of the applications an operator wants to host in that cell, Chan said. Operators may choose to start off simply with some simple caching features at high-traffic locations, requiring less powerful processing capabilities. But as they evolve their network topologies to support huge overlays of small cells, they’ll likely push more intelligence to the edge of their networks, turning the cell site into the equivalent of a small data center. As that happens, operators will need some real processing power, Chan said.
Want to use this article? Click here for options!
© 2014 Penton Media Inc.
From the Blog
Join the Discussion
Get more out of Connected Planet by visiting our related resources below:
Connected Planet highlights the next generation of service providers, as well as how their customers use services in new ways.Subscribe Now