The Coming of Cloud RAN

By | January 6, 2014

Cloud RANThe evolution of wireless communications has spawned many innovations but the cellular concept where base stations are dispersed to meet coverage and capacity requirements remains constant. This is about to change, sort of, if Cloud Radio Access Network (CRAN) architecture lives up to its vision. What makes CRAN such an interesting development is that it leverages advances in a number of different fields to pull together a solution to two perpetual problems facing network operators: cost reduction and increased capacity.

Cloud RAN is much more than simply centralized baseband modems connected by fiber optical cable to remote radio heads (RRHs) – this is commonly known as base station hotel or centralized RAN. CRAN, on the other hand, comprises a virtual pool of baseband units (or digital units [DU]) which drives any RRH in the pool. The baseband processing runs on commercial servers which is in itself a challenge in process of being resolved by offloading certain computationally expensive processes in the physical layer to an accelerator processor. For those who grew up with the IBM PC and Intel’s early processors such as the 8086 would recall the math co-processor option installed to speed up computational tasks. Virtualization and pooling allows the network to switch capacity (physical resources) from a light-traffic area to a heavy-traffic area. Running base station function on commercial servers in a data center is of great strategic significance as it alters the balance of power between operators and vendors in favor of the former. The theory is that operators with standard commercial hardware are free to select any vendor for the software. Should this vision be realized, the ecosystem as we know it today will be shaken to its core.

From a business case perspective, CRAN trades the cost of fronthaul fiber (the connection between DUs and RRHs) for reduction in equipment cost, site rental fees, energy consumption, and network operation and management expenses. It also works to speed up deployment of new sites because less equipment is placed at the cell site. But the business case varies widely between operators and regions. Where and how CRAN will be implemented is something that I explored in detail in a report (Cloud RAN – A market analysis of radio access network evolution) published recently by Mobile Experts and Xona Partners which includes case studies comparing CRAN deployments in different geographies and deployment scenarios validated by major operators on three continents. It goes without saying that the availability to fiber is prerequisite for CRAN deployment. But this is part of the story. There are certain deployment scenarios where wireless fronthaul will play a role.

One such scenario is CRAN in HetNet deployment where low power RRHs are used instead of compact base stations (which has become synonymous with small cells, but I believe this is a misleading nomenclature). If the cost of deploying a compact base station or low power RRH is driven by the logistics of the process, I don’t see much difference in the overall cost between the two scenarios. Even as fronthaul requires Gbps capacity whereas backhaul requires 100’s of Mbps capacity, the difference in the cost of equipment would pale relative to the other costs. This makes CRAN a direct competitor to outdoor compact base stations. The advantage that CRAN brings to this scenario is that tight coordination between the macro cell and small cell layers brings about an increase in site average capacity on both the uplink and downlink path.

For CRAN HetNets to become a reality, wireless fronthaul needs to meet stringent requirements for high capacity (order of Gbps) and latency (order of tens of microseconds) as one cannot expect fiber to be available at any location. Such requirements mean that E-band systems (70/80 GHz) and to a more limited extent V-band systems (60 GHz) will play a major role in implementing the CRAN HetNet architecture. Optimization techniques to limit CPRI line rate between baseband and RRH will contribute to increasing the utility of wireless fronthaul solutions. There are many efforts on this front today.

The roadmap to LTE-Advanced is full of features that squeeze capacity out of the network by implementing intelligent algorithms that coordinate transmissions between different base stations (CoMP, eICIC, network MIMO, coordinated scheduling, etc.). Realizing the full potential of LTE-Advanced relies on Cloud RAN-type architecture where tight coordination between baseband resources is possible. Cloud RAN in turn is spawning a number of developments that include baseband architecture, techniques for pooling and virtualization of baseband processing, fiber optical networking, timing and synchronization and wireless fronthaul. For this reason, Cloud RAN is one of the most interesting developments to occur to the base station and network architecture. It is set to be a focal point of industry attention in the next few years.

Note: The 75-page market study investigates the costs and benefits of Cloud RAN:

  • Full descriptions of how Centralized RAN (Base Station Hotels) and Cloud RAN will evolve;
  • Case studies to quantify the benefit of CoMP and eICIC;
  • Estimates of the cost savings and cost increases with different CRAN deployments;
  • Cost comparisons for CRAN vs DAS deployment;
  • Monte Carlo analysis to show the sensitivity of the CRAN business model;
  • Specific details on how CRAN compares to DAS networks for coverage and capacity;
  • A five-year forecast of adoption for CRAN;
  • A supporting Excel spreadsheet with detailed cost and forecast analysis.

For more information click here.

6 thoughts on “The Coming of Cloud RAN

  1. Kit Kilgour (@kitkilgour)

    Good article. I see that the recent China Mobile report on their CRAN trials with some of the major vendors indicates that standard GPUs cannot yet do all the low level LTE processes such as HARQ. It makes me wonder whether some recent purchases by Intel are intended to help address this via some co-pros

    1. Frank Rayal

      Definitely the acquisition of Mindspeed will provide Intel with added muscle in LTE protocol stack development.

      1. Kit Kilgour (@kitkilgour)

        Mindspeed do parts of the physical layer and silicon only, not higher layers of the stack, so the protocol stack gains might be limited – but perhaps some firmware and hardware enhancements

  2. Pingback: Is the Personal Cell Technology for Real? | Frank Rayal

  3. Pingback: Is the Personal Cell Technology for Real? | ytd2525

Comments are closed.