Intel’s wireless ambitions go beyond smartphones and tablets. It’s set its sights on the guts of the mobile network as well. By embracing a new network design concept called Cloud-RAN, Intel believes it can reshape wireless networks from highly-specialized architectures into more generic computing platforms that run over its off-the-shelf silicon. And in the world’s largest operator, China Mobile(s chl), Intel sees the opportunity to make that vision happen.
China Mobile has a massive network of 700,000 GSM and 220,000 3G base stations built into towers throughout China’s vast landscape. The base station is easily the most expensive element of the wireless network, and as China Mobile looks to the next wave of wireless technology, LTE, it doesn’t want to repeat that enormous infrastructure investment by installing pricey hardware at the bottom of every tower. Instead, it’s looking for Intel’s help to move all of that network intelligence into the cloud, leaving only the radios and antennas at the cell site.
The Cloud-Radio Access Network (Cloud-RAN) isn’t the public cloud of Amazon Web Services. Rather it’s a private cloud run by each operator in local data centers, but the principle is the same. China Mobile could centralize an enormous number of now-distributed computing resources. That would not only save capital and operating costs, but it would also allow it to webscale the network’s biggest number-crunching requirement – converting the analog fuzz scooped out of the airwaves into digital ones and zeros the network can understand.
Using supercomputing principles to handle baseband processing means no longer having to build networks to meet peak demands at every tower. Cell sites usually see huge upticks in use at a few predictable times each day: during work hours in a business district, for instance, and mornings and evenings out in the suburbs. Outside of those peak times, that capacity just goes to waste. But with Cloud-RAN, operators can allocate capacity where and when needed, following the flow of network congestion from the suburbs to the central city and back again. By putting their base stations in the cloud, operators could drastically cut the processing power necessary to run the network as a whole – by some estimates as much as 40 percent.
Meet the Cloud-RAN players
Alcatel-Lucent(s alu), Nokia Siemens Networks (s nok)(s si) have developed Cloud-RAN platforms of their own, giving them fancy names like lightRadio and Liquid Radio respectively. Meanwhile, chipmakers Texas Instruments(s txn) and Freescale(s fsl) have both begun retooling their baseband designs for future cloud implementations. But Intel aims to take the concept one step further. Instead of merely relocating base stations to the cloud, Intel proposes using its multi-purpose Xeon processors to perform the same signal processing tasks that are now the purview of highly-specialized equipment.
In short, Intel wants to replace the big-iron wireless networks of today with what are essentially server farms that can be built and deployed for a fraction of the cost.
At a telecom industry conference last year, the GM of Intel’s Communications Infrastructure Division, Rose Schooler said that the telecom industry has been hobbled by its fixation on proprietary network interfaces, opaque platforms and a morass of complex signaling protocols – that’s code for telecom vendors not wanting to open up any more than wireless standards require. By adopting the more open standards of the computing industry, the wireless industry could innovate at the faster pace of computing, Schooler said.
Not surprisingly, the traditional telecom vendors think Schooler and Intel suffer from a case of wishful thinking. Freescale and TI have also tapped into computing architectures for their next generation of chip designs. Freescale uses IBM’s PowerPC, while TI has begun integrating ARM cores into its latest basebands. But TI hasn’t done away with the key proprietary component of its baseband designs: the digital signal processor (DSP). No matter what Intel claims, wireless networks are highly specialized creatures and therefore require highly specialized silicon, said Tom Flanagan, director of technical strategy for TI’s wireless base station infrastructure team.
“Its kind of naïve to think that you can replace this highly optimized technology with something general purpose and not lose anything,” Flanagan said in a recent interview. “We build that expertise into the hardware because hardware is exactly where it needs to be.”
For Intel to replicate the base station with generic Xeon processors would require it to build many of its functions into software, and doing baseband processing through software is a highly inefficient way to run a network, Flanagan said.
What are Intel’s chances?
To build a wireless network business, Intel doesn’t just need to battle the established telecom vendors, it has to make the case for Cloud-RAN to the wireless carriers. Last year at CTIA, Verizon(s vz) CTO Tony Melone was dismissive of the new cloud architectures emerging at the show, saying that they were neat design concepts, but hardly ready for prime time. Verizon Wireless is the world’s most aggressive carrier when it comes to LTE and implementing new network technologies, so Melone’s lack of endorsement is telling.
Another obstacle is the enormous backhaul capacity that a Cloud-RAN architecture would require. Sending raw unprocessed radio frequency data over the network to a cloud data center would require much more bandwidth than a copper or microwave backhaul link could provide. That means fiber is the only way to support Cloud-RAN, and no operator has fiber links to all of its cell towers.
But if Intel lands a contract with China Mobile it may not have to worry about other customers. An LTE deployment from China Mobile could eventually scale to more than a million cells, translating into a heck of a lot of high-end chip sales. Intel, however, faces competition from the incumbents on that front as well. Intel has been working with on China Mobile’s Cloud-RAN project since its inception, but recently the carrier began testing Alcatel-Lucent’s lightRadio technology as well.
Given the decline of its core PC business, Intel needs to find new markets for its X86 processors and figures that the wireless industry is ripe for the picking. But its attempts to break into other aspects of that industry have flopped. Intel has tried for years to challenge ARM’s dominance in the mobile computing market with its Atom processors, but the X86 architecture’s notorious power problems have kept handset makers disinterested. Only earlier this month did Atom start showing life. Most recently Intel has been trying to position its processors as a means of making dumb femtocells and picocells smart.
I’ll give Intel one thing. It’s tenacious. With every wireless initiative failure – note WiMAX – it immediately launches another. And where it can’t develop a wireless technology on its own, it buys a company that can. Cloud-RAN could be a key turning point for Intel or another flop, but we’ll probably have to wait several years to find out. While China Mobile is trialing LTE now, its commercial rollout could be as far away as 2014.