
Cloud RAN and Mobile Edge Computing, a dichotomy in the making
Caching content locally based on user preferences for lower latency and handling ephemeral data such as location based analytics needs edge computing. These two architecture concepts are proposing to deploy compute at different nodes within the network. On the surface, these two competing architectures appear to be creating opposite pulls, resulting in a dichotomy within the network. A deeper look suggests that a balanced approach to deploying networks could leverage merits of both, thereby transitioning these two competing technologies to complement each other for enabling new services.
More than a decade ago, the concept of distributed base stations emerged with a desire to overcome power loss in sending signals using coaxial cable from traditional base stations located at the foot of towers to antennas mounted on the top of tower. Radio heads were located remotely in close proximity to the antenna on tower top to eliminate power losses. Remote radio heads were connected to baseband BTS chassis using fiber. Protocols such as the Common public radio protocol (CPRI) were devised to transport data and synchronize remote radios. In some situations when fiber was not available, microwave or millimeter wave radios were used to transport CPRI payload. This architectural shift raised hopes for operators to mix and match radio and baseband chassis from different system vendors to lower costs, improve supply chain, and ease inventory management. Interoperability concerns prevented this from happening, nevertheless it opened the way for tier-1 system vendors to leverage radios from smaller vendors for managing the rapid increase in a variety of radios for different geographies.
Distributed base station architecture has taken roots. This architecture is centralizing the baseband processing pool, sometimes called super Macro, which is capable of feeding a larger number of radios and therefore more effective coverage and load balancing. Success of data center and Cloud computing resulted in the emergence of Cloud RAN concept that extends distributed base station architecture by virtualizing base band pools running on server farms. There are merits in Cloud RAN as it allows use of lower cost compute, leveraging off-the-shelf server chassis for cost effective RAN deployment, load balancing and significant ease in network provisioning. Cloud RAN, when implemented broadly, holds the promise of allowing third party providers to own the network, enabling multiple virtual network providers to concentrate on content and services.
Cloud RAN architecture is seeing some early acceptance in the Asia pacific region where operators have significant fiber assets to deploy remote radio heads. Carriers are investigating hosting layer 1-3 base station stacks and evolved packet cores on off-the-shelf servers as virtual machines. General purpose compute cannot implement the layer 1 baseband functions, packet processing, and security efficiently for high throughput and low latency. These functions would require servers to use specialized accelerator cards. The ability to host base stations as a set of software functions offers significant benefits. Carriers no longer need to build out network gear per peak capacity requirements. Instead, base stations can be instantiated in the Cloud on a need basis to provide the desired coverage and capacity. Cloud RAN allows base stations to be co-located in data centers where most of the content resides. This leads to higher efficiencies and effective dissemination of content.
There are few hurdles on the way to Cloud RAN that are slowing down adoption. Latency and low jitter long distance connectivity to remote radio heads is a big challenge. Off-the-shelf servers do not have compute resources to efficiently run baseband processing. Telco grade servers with accelerator cards for layer 1 baseband are needed to host pools of baseband processing running in virtualized environments. System vendors that are lagging in certain geographies are championing this cause to disrupt markets and gain market share, forcing incumbents to follow suite to secure their market share. Carriers are welcoming this trend with a desire to harmonize their Cloud computing assets with network infrastructure to ease deployment and maintenance.
Figure 1: Cloud RAN network architecture. Click image to enlarge.
Distributed base stations have their own unique benefits in terms of caching content as per local users’ preferences for improving service delivery and processing data close to the source for latency sensitive applications. Proximity to users at the edge results in ultra-low latency access that opens opportunities to deploy customized services. MEC envisages convergence of IT and communication segments at the network edge to enable new services and business segments. Location services, Internet-of-Things (IoT), video analytics, augmented reality, local content distribution, and data caching are some of the use cases identified by MEC. MEC architecture proposes adding servers to macro and super macro base station sites for local compute and storage to enable new applications. Application development stack, tools, and framework are in the making to allow ecosystems to launch new applications and integrate services for multiple business verticals. Key hurdles in the path of MEC are cost in terms of space rental for adding server and storage to base stations, maintenance, and charging policies. Currently, policy charging and rules function is part of the core network controlled by carrier. A derivative PCRF function would need to be hosted locally at the base station to allow carriers and other content providers to fairly charge end users for services.
Figure 2: Conceptual diagram of Mobile Edge Computing. Click image to enlarge.
5G proposals for the information society of 2020 are adding further twists to the dilemma imposed by the architecture pull between Cloud RAN and MEC. In order to meet projected data demands in 2020 with increasingly scarce and limited spectrum, 5G proposals are aiming to continue to improve spectral efficiencies using techniques such as Massive MIMO for both <6 GHz and >6 GHz spectrums. Massive MIMO systems use large number of antennas to create beams per user. This allows significant improvement in energy efficiency and throughput. An additional benefit of massive MIMO is use of inexpensive lower power components for antenna signal chains. Massive MIMO techniques are well suited for millimeter and centimeter frequencies, an inexpensive and underutilized spectrum resource that is available in large contiguous chunks. Narrow pencil beams at these frequencies result in large antenna gains that compensate for high propagation. Along with these benefits come some hurdles. There is a significant increase in complexity in supporting a large number of active radio signal chains and layer 1 baseband with pre-coding for digital beamforming. Bandwidth requirements increase sharply between baseband processing signal chains and radios. In order to economically realize these systems, it is necessary to integrate layer 1 baseband signal processing with the radio. Such a functional split in the future may lead to network nodes that may go back to traditional base station architecture where all L1-L3 and radio functions are co-located.
Figure 3: Millimeter wave massive MIMO (200 MHz, 64×64 antenna array) system. Click image to enlarge.
Mobile edge computing and massive MIMO techniques may imply consolidation of distributed base stations thereby hindering the move to Cloud RAN. In reality, limited spectrum necessitates leveraging merits of multiple different network architectures co-existing together to meet an ever increasing growth in demand for bandwidth. Cell densification enables reuse of a scarce spectrum resource. In coming years, this trend will result in distributed base station sites to become more like mini data centers. On the other hand, splitting Cloud RAN into numerous mini data center may be a way to overcome tight deterministic latency and synchronization requirements with remote radio head connectivity. The two architectures may appear to converge in the middle. Cloud RAN and MEC architectures can coexist to complement each other. Cloud RAN can rely on latency and the proximity merits of edge computing nodes and edge computing can benefit from centralized network deployment, management, and service provisioning. Only time will tell how these two architectures would see adoption in the next 3-5 years as industry gears up for 5G deployment. End user applications, operator preferences based on cost of network deployment and maintenance, and system solutions from equipment vendors may likely be key factors in dictating a delicate balance between the two.
In summary, heterogeneity in wireless network is expected to continue to increase. There is unlikely to be a clear winner. A good balance of Cloud enabled RAN and mobile edge compute equipment is needed to effectively serve wireless broadband services. Instead of letting the hype pendulum swing to one extreme, the broadband wireless ecosystem needs to make balanced investments to continue to build complimentary technologies to effectively serve the information society of 2020.
