A question of efficiency: C-RAN and SDN

As an acronym, C-RAN has less name recognition than its more popular cousins, SDN and NFV, though if you are in the wireless business it is equally important as we look at the future of the wireless network and its continued evolution over the next five to ten years. However, C-RAN has not been widely implemented to date due to challenges resulting from the lack of efficiency in using general hardware to do specialized hardware’s job.

C-RAN can stand for cloud RAN and other times for centralized RAN, which can be considered two sides of the same coin. In either case, C-RAN reduces maintenance by concentrating base station resources in a single location and reducing power through the use of remote radio heads. Cloud RAN, which is what we usually mean when we say C-RAN, means that the network is virtualized. There is a generic compute server or blade, and base station resources can be added as wanted or needed to support the network traffic. It runs on a standardized server infrastructure and typically uses general purpose processors. On the flip side, centralized RAN features more dedicated components within a blade architecture. It can be easier to implement, but offers fewer of the benefits associated with virtualization, including reduced maintenance. Given the purpose-built hardware approach to centralized RANs, they will arguably consume less power.

The more well-known acronym, software-defined networking (SDN), enables the separation of the control plane and the data plane, which can also be applied to C-RAN environments. It enables universal remote management capabilities, allowing the operator to remotely manage the network, comprised of base stations, C-RAN, small cells and other components. Network Function Virtualization (NFV) enables the implementation of virtual services through the virtual machines on the general purpose processors, allowing network operators to push new applications and services to the network edge. So if you choose to implement a C-RAN environment, applying the principles of SDN or NFV, most of the processing is implemented in a general purpose compute environment, allowing new virtualized functions to be added to the network.

This is where value-added operator services could be added to the network. For example, picture yourself at the World Cup, which starts this week in Brazil with an expanded network supported by us (see Bend it with base stations). Imagine 80,000 or more fans packed into a single stadium, and an amazing play happens, resulting in a goal. The first thing several thousand fans are going to do is pull out their smartphones to access the instant replay. That video will replay several thousand times across several thousand devices in short succession, placing a big strain on the network. However, in a C-RAN enabled general purpose compute environment, with value added services enabled through SDN and NFV principles, the server might store a local copy of that replay video so that each request is not sent all the way back through the core network. This reduces lag time and ensures that each of those several thousand viewers get instant access to the instant replay. This is known as content caching.

In addition to content caching, you may also be familiar with some of the load sharing capabilities of SDN. These same features are echoed in a C-RAN environment. A typical network must be built so that the base stations can handle the peak level of traffic at all times. However, over the course of the day, base stations are at best 20 percent utilized during off peak hours. C-RAN allows the aggregation of those base station resources and the virtualization allows for load sharing. If a single set of base station resources covers both a residential and commercial areas, for example, resources can be allocated appropriately — supporting more traffic in residential areas before 8 am and after 5 pm, while enabling increased resources in the commercial area during business hours.

As always, of course, this does involve some compromise. Specialized hardware, which would typically not be used in a C-RAN environment, is more efficient than general purpose processors. Therefore, the benefits achieved through the load sharing or value-added services must be greater than the added cost for more general purpose processors needed to support the network.

When you get down to a real-time implementation, we believe a hybrid approach seems most likely. Layer 1 is where the majority of the wireless processing of the physical layer in base stations occurs and is the most compute-intensive layer. With dedicated hardware for Layer 1 processing and a shared pool of general compute resources for Layers 2-7, the benefits of specialized hardware’s efficiency can be realized, while still taking advantages of the virtualized features of C-RAN.

Jeff Steinheider is a product marketer in Digital Networking for the QorIQ Qonverge family of SoCs and Wim Rouwet is senior systems architect — distinguished member of technical staff.

Jeffrey Steinheider
Jeffrey Steinheider
Jeff Steinheider is a Product Marketer in the Digital Networking business group, where he works closely with customers to ensure that their design needs are met. He holds Masters and Bachelors degrees in Computer Science and Electrical Engineering from MIT. Jeff enjoys playing soccer with his kids and downhill skiing—though he’s still trying to find some snow in Texas. Find him on Twitter at @jlsteinheider.

Comments are closed.

Buy now