Intel's bet on Cloud-RAN vs traditional RAN proccessing

Intel (and a few other vendors) believe the future of the telecom architecture is moving the radio processing element from base stations to the cloud (i.e. operator's datacenters). They propose using general purpose processors like Xeon for processing baseband data rather than propriety ones. Distributing the processing grunt work across server farms would in theory provide cost savings as well.

source: #-Link-Snipped-#

So, where do you guys think the signal processing should lie? I dont think Intel's cloud-RAN isnt a good approach.

Firstly, Xeon processors are not optimized for signal processing of radio data. I'm all for the concept of software defined radio (which incidentally can be done at BS side too), but not when general purpose processors are used which are highly inefficient for radio processing.

Secondly, sure, it costs a lot to install base stations at every node for now, but you save significantly on bandwidth. Per node, Raw radio data can consume 3-4 Gbps per node before being processed down till 100 Mbps (in most cases) before transferring across the A Interface/R6/S1 data link. If processing is offloaded to server farms, a cluster of nodes would require significant fiber back haul resources already!

I believe costs on RAN equipment can decrease once vendors combine hardware resources at base station side. New DSPs are always coming out, being more efficient and require less hardware resources. If cloud computing is to be leveraged, its better to still handle the heavy radio processing on basestation but offload other tasks to the central cloud.

I suppose Intel just wants to be part of the action. Unfortunately they'll need to do just more than offering Xeon based cloud solutions. Haven't they learned anything yet from their battles with ARM in the mobile segment? 😉