Evaluating processing options in RAN: Purpose built and Cloud RAN
RAN Evolution Expert
Strategic Product Manager, Cloud RAN Portfolio
RAN Evolution Expert
Strategic Product Manager, Cloud RAN Portfolio
RAN Evolution Expert
Strategic Product Manager, Cloud RAN Portfolio
Introduction
With the advent of technologies such as Cloud RAN and the specifications developed in the O-RAN alliance, service providers are faced with questions relating to new and different options to process RAN functions:
- Is it best to execute in a de-centralized way, close to the antenna, or in a more centralized location?
- How is purpose-built hardware and cloud infrastructure best combined for RAN processing?
- How can total cost of ownership (TCO) be optimized across the whole system?
A new processing baseline for 5G
Before addressing the above questions, let’s look at service providers two key expectations on 5G:
- Manage the anticipated traffic growth of 4X from now until 2025 as predicted in Ericsson Mobility Report
- Leverage mobile networks to support new use-cases such as critical IoT, fixed wireless access, Ultra reliable low latency connectivity etc.
To meet these expectations service providers are building 5G networks by:
- Adding new and wide frequency bands, i.e. mid band and high band.
- Deploying advanced antenna systems with massive MIMO for additional coverage and capacity.
- Using shorter time transmission interval (TTI) to lower latency.
These expectations and actions impact and increase the requirements on the way service providers design their 5G networks and decide on the best RAN processing option.
The 5G RAN protocol stack
Now let’s turn our focus to the 5G RAN protocol stack to outline the different parts where RAN processing is important. 3GPP has defined a Higher Layer Split interface (F1), separating the Central Unit (CU) from the Distributed Unit (DU). In summary, the 5G protocols contain:
Figure 1. 5G RAN protocol stack
The lower you go in the protocol stack, the higher the demand on processing – Layer 1 & 2 combined comprise 90% of processing demands. The wide 5G mid and high bands together with massive MIMO technology exponentially increase the processing demand on L1 and beamforming (Figure 2). In addition, the transport overhead increases the lower you go in the protocol stack and for this reason the de-centralized processing has the advantage of lowering the requirements on RAN transport.
Figure 2. Processing demand per protocol layer
Processing in RAN is diversifying
When 4G was built networks were homogeneous with a de-centralized baseband on every antenna site. Since 5G RAN is more heterogeneous with a combination of low, mid and high bands, the processing is better suited to be diversified to meet different needs. Below is a list of possible set ups (non-exhaustive) :
Figure 3. Illustrative deployment set ups and service provider value. (legend as in Figure 2)
Processor options in RAN
Purpose built hardware, such as Ericsson RAN Compute Baseband, uses system on a chip (SoC) which is a tailor-made chip based on Ericsson Many-Core Architecture (EMCA) for lower protocol layers and uses X86 processor for the higher layers.
X86 processors are used in Cloud infrastructure today. Ericsson has shown in a live demo that the complete stack for low band 5G is feasible to process on such infrastructure. For mid band and massive MIMO, hardware acceleration will be needed as a complement for the demanding L1 processing.
Now let’s look at the merits of de-centralized and centralized processing for both purpose-built and cloud infrastructure.
De-centralized processing
- Purpose-built hardware - For de-centralized processing on the antenna site the purpose-built hardware can integrate into massive MIMO radio for beamforming and full Interference Rejection Cancellation (IRC)-receiver processing, into routers and gateways for CPRI to eCPRI conversion and into basebands and outdoor radio processors.
- Cloud infrastructure - If rack space is available on the antenna site both purpose-built and cloud infrastructure processing is feasible.
Centralized processing
- Purpose-built hardware – Purpose-built hardware has the advantage of maximizing the compute power in a sub-rack resulting in a more compact installation.
- Cloud infrastructure - For centralized processing on cloud infrastructure the advantage is that synergies can be found when multiple applications such as RAN, core and content/compute can share infrastructure resources in the cloud.
Ericsson Radio System and 5G cloud core is the foundation
Ericsson has more than 100 5G commercial agreements, all built on a foundation of Ericsson Radio System (ERS). It is powered by a tailor-made system on a chip (SoC) with special purpose processors, based on Ericsson Many-Core Architecture (EMCA). EMCA is future proof and optimized for maximum performance with a small form factor and low power consumption. It enables integration into many Ericsson Radio System products including massive MIMO radios, routers and RAN Compute allowing for a coherent and integrated system performance proven in the field.
It is less attractive for service providers to invest in cloud infrastructure for 2G/3G/4G since this is already deployed and a mature technology handled by multi standard, purpose-built hardware. In addition, Ericsson Radio System radios and RAN compute portfolio supports 5G with a remote software installation.
Service providers now deploy 5G Core based on cloud infrastructure, capitalizing on Cloud native technologies for a more efficient and TCO optimized deployment. As service providers evolve this cloud infrastructure it is important to consider the specific demands of 5G RAN use cases so that the infrastructure can be shared across domains (e.g. running both 5G Core and RAN).
Ericsson provides the possibility to complement purpose-built with cloud infrastructure
Ericsson Radio System interfaces evolve to also enable RAN processing on cloud infrastructure:
- Massive MIMO radios powered by EMCA SoC for the heavy processing in L1, maximizing spectral efficiency and providing a packetized fronthaul interface (eCPRI). The packetized fronthaul enable processing also on cloud infrastructure.
- EMCA SoC is used in routers to implement L1 CPRI to eCPRI conversion for remote radio units. The packetized fronthaul enable processing also on cloud infrastructure.
- High band street macro with EMCA SoC for the lower layers can benefit from a common anchor point with interconnection to the core. This anchor point can be realized as a central unit (CU) on cloud infrastructure, with potential to manage load balancing more holistically with new scaling capabilities.
Processing options in 5G core networks are also diversifying. Core network processing has traditionally been done on few centralized sites. But there are benefits with more de-centralized core processing such as low E2E application latency and Security. Such de-centralization of 5G core brings cloud infrastructure into edge or far-edge data centers creating processing synergies with the RAN since these new cloud sites can also process RAN functions.
Ericsson foresees a stepwise introduction of cloud infrastructure for RAN processing to complement Ericsson Radio System. Here are three examples:
- Introducing unique interworking features such as Ericsson Spectrum Sharing and Carrier Aggregation across purpose-built RAN and cloud infrastructure.
- Centralized RAN deployments in strategic locations can serve many cells provided that sufficient transport capability is in place. This type of deployment enables better resource pooling, efficient scaling capabilities on cloud infrastructure and potentially also removing the need for hardened server deployments on de-centralized sites.
- Enterprise and industry edge are use cases that could benefit from RAN deployed on cloud infrastructure by simplifying operations & maintenance, enabling local data sovereignty and use of Machine Learning/Artificial Intelligence applications.
Check points for evaluating processing options in RAN
As a conclusion there are some technology agnostic check points that are generally applicable for service providers when evaluating processing options in RAN.
Evaluate 5G RAN processing:
- to cater for high load to support the expected traffic growth
- to ensure support for future use-cases
- in combination with transport
- considering pooling gains of centralized vs. de-centralized
- to optimize full RAN and transport TCO
- quantifying the value in operations and maintenance by harmonizing onto uniform COTS hardware
- taking into consideration the potential co-location synergies of core break-outs, far edge- and centralized RAN-deployments
We are in the beginning of an exciting journey with a multitude of RAN processing options that will meet the expectations of managing the anticipated traffic growth and support new use cases. Ericsson is convinced that a complementary approach of purpose-built and Cloud RAN processing is the most ROI friendly approach for service providers. This as it gives a broad toolbox to realize benefits in a variety of deployment scenarios.
Watch the Evaluating processing options in 5G RAN video
Read more about Core network
