Will the real CRAN please stand up?
April 2016
The mobile industry loves its acronyms, like a dog loves a bone. We chew on them until they’re ragged and shapeless. CRAN is a good example. The original meaning of the term has been lost in a flurry of marketing hype.
Cloud RAN? Centralized RAN? Coordinated RAN? Collaborative RAN? Clean RAN? What does it all really mean?
“Centralized RAN” is well underway. Verizon, NTT DoCoMo, SKT, and other operators have deployed base stations with more than 700,000 radios in the field, tied to baseband processing in a centralized location. So far, most of the radio heads use a CPRI interface which works best on fiber, with high data throughput and low latency/jitter.
“Coordinated RAN”, also called “Collaborative RAN”, is evolving. Once the RAN is physically deployed in a centralized configuration, multiple radio heads can work together to improve the data speed for users and reduce interference in the system. We’re still in the early days of LTE-Advanced, and software upgrades over the next 5-10 years will result in capacity and throughput improvements. In spots with poor coverage, data speed will double when “Coordinated RAN” comes into play.
“Clean RAN” is a marketing slogan, simply indicating an environmentally friendly network with high energy efficiency.
Nobody really meets a strict definition of “Cloud RAN” today. In cloud computing, data is stored centrally and software runs as a “virtual machine”, on a standard off-the-shelf server. Full virtualization means that the entire process can be expressed in software that runs on a generic server. Virtual machines are easy to scale up and the maintenance is done by technicians in a large server farm, so they can specialize and run the servers efficiently. Amazon Web Services, Microsoft Azure, and others have built huge server farms….and the resulting cost savings has fueled a big section of the tech industry for the past five years.
Operation on a true generic server hasn’t been proven in a full-scale LTE network, although Ericsson has shared some impressive results through NGMN. Our interviews with big OEMs and operators indicate that some basic technical hurdles remain, and hardware accelerators are likely to be added to modify the server.
If the OEMs actually introduce RAN software that runs on generic servers, it’ll be a big win for the operators. Operating costs can drop dramatically and new levels of competition will drive cost down in general. For multiple reasons, I think that the hardware is likely to be modified, where the X86 processor handles the heavy lifting but hardware accelerators take a few key steps to make it work. Read our CRAN research if you want all the techie details.
The small cell vendors that offer “Cloud RAN” products are really offering variations on Centralization and Coordination, without even attempting to run on generic servers. In several cases, baseband processing is split so that radio-related processing takes place in the radio node and other functions are centralized in a hub that resides within a hundred meters of the radio node. It’s not the original concept of Centralization but the same benefits can apply. Small-cell implementation such as the Commscope OneCell can get those benefits without the high cost of macro OEM software.
If the big OEMs follow the pattern of history, they won’t introduce truly interoperable products. They’ll keep some proprietary features which tie hardware and software together. There are legitimate technical reasons for this, but frankly it can also serve to keep barriers to entry in place. This approach would be pretty similar to the development of OBSAI and CPRI in the early 2000’s. CPRI didn’t end up with true interoperability at the box level, but by defining CPRI the industry was able to scale up the Serdes devices and FPGAs that were necessary, and the same semiconductor devices were used by several OEMs.
Getting close to a true Virtual Machine is good, but it’s not the same as getting there. If the conclusion is that a “virtualized RAN” means a modification to the off-the-shelf server, the operators won’t get open competition and commoditization. It’s time for the operators to push, in a concerted effort, to determine the feasibility of generic hardware. NGMN is set up to do this, and every LTE operator should support their activity in this area.
In short, most of the capacity benefits from LTE-A will come from Centralized RAN. The additional benefits of operational efficiency will come from virtualization. We may not be able to achieve full virtualization, with all of the cost benefits of “The Cloud”. Even if each operator has its own private Cloud, then each can benefit from some efficiencies in operations and maintenance.
It’s ironic that, to maximize the savings of the Cloud, mobile operators need to set rigid requirements.
Comments
Post a Comment