As companies increasingly start to build and run their own 5G networks, there is much talk about open standards in cellular radio access networks (RAN). The rise of the Open RAN standard allows vendor interoperability along with virtualized RAN on open, cloud-native hardware, both at the near and the far edge.
Historically, a single vendor provided the bulk of a mobile carrier’s infrastructure in a given market, including base stations, remote radio heads, controllers, gateways and the core network. Mobile networks today are moving to a disaggregated model that reduces traffic and emphasizes locality of data processing. This trend benefits operators who want greater interoperability and flexibility in today’s 5G network architecture, between distributed units (DUs) and radio units (RUs), and between DUs and centralized units (CUs) — all from multiple vendors.
Of course, disaggregation is not without risk. Open, multi-vendor mobile systems need to provide feature and performance parity with existing networks, including a network reliability goal of 99.999% uptime. The good news is that silicon innovation, coupled with optimized software solutions, is allowing systems integrators, and in some cases, network operators to step up to this integration challenge in the new multivendor O-RAN environment.
But why did disaggregation happen in the first place? The main reason for the shift was the rise of 5G and its immense processing complexity at the network edge. If you look closely at this disaggregation, you see it is happening at two levels: CU to DU, and DU to RU. This is necessitated by a significant increase in the amount of bandwidth that is required.
Conventional architectures for 4G/LTE base stations typically operate at 20 MHz of single channel bandwidth with 2-to-4 antennas. 5G TDD spectrum channel jumps to 100 MHz and is increasingly deployed with 32 to 64 antennas in a massive MIMO array. This massive increase in bandwidth and antenna capacity results in a 50x greater bandwidth requirement than a typical 4G case. This is well beyond the capacity limits of a typical fronthaul interface.
One direct consequence of this is more intelligent processing in the radio head (aka RU). This is a dramatic shift because radio heads are historically analog devices with minimal control and logic. This also has necessitated a new interface between the RU and the DU. The choice of standard Ethernet for this new fronthaul also resulted in providing operators flexibility in L1 processing and the ability to procure RUs from any vendor they choose, and that those RUs will work with DUs from any other vendor they choose. The O-RAN alliance standard codifies the methodology and protocols for this new fronthaul.
One of the key benefits of O-RAN, particularly in what is known as the 7.2 split, is a partitioning of the lower and upper physical layer (L1) processing between the RU and DU respectively. This enables the RU to order to efficiently consolidate approximately 300 Gbps of raw data, discard the excess and convert the rest into a 25 Gbps stream before sending it over the fronthaul link to the DU. The DU then handles the upper L1 processing and forwards the data to the upper L2/L3 layers.
To be competitive, O-RAN systems need to be at feature parity with traditional highly-integrated RANs. This avoids compromising essential features and performance for virtualization and vendor interoperability. They also need to be truly open in a way that does not require an instruction set tied to a vendor-specific high-end CPU. A highly-integrated Arm-based O-RAN SoC optimized for 5G use cases with programmable cloud-native acceleration allows deployment at scale in leading carrier networks. These solutions uniquely provide highly integrated in-line O-RAN L1 acceleration, as well as industry leading L2/L3 processing both for conventional and virtualized architectures.
For the industry, this enables 5G networks to live up to demanding performance expectations in new industrial use cases from factory automation to smart cities. By taking a future-ready approach, such solutions are also giving rise to self-optimizing radio networks with ultra-low latency AI/ML in-line acceleration on the same platform. These advantages promise to scale connection quality and reliability to new heights for all users, including billions of mobile consumers worldwide.