Optimal Compute Architecture for vRAN/Open RAN Solutions
Written on
Chapter 1: Overview of vRAN/Open RAN
As the telecom sector gears up for the Mobile World Congress, it's essential to reflect on the current state of the vRAN/Open RAN market and the pivotal trends, particularly in compute architecture. It is becoming increasingly evident that success will favor those companies offering advanced technologies, competitive features compared to legacy RAN, and solutions that are both power-efficient and cost-effective.
When operators initiate their transition, it appears to be a two-phased strategy. The first phase involves implementing a single-vendor vRAN with open interfaces, followed by a second phase where multi-vendor open RAN systems are deployed. This strategy not only reduces the complexity of system integration but also facilitates a smoother transition. Operators are beginning to realize that the anticipated cost reductions associated with open RAN may not be as substantial as initially believed. In fact, early deployments might incur higher costs. However, the elimination of vendor lock-in during the second phase could lead to future savings.
Feature parity with established 5G networks is another significant consideration. While initial vRAN/Open RAN deployments primarily utilized simpler configurations like 4T4R and 8T8R MIMO, more sophisticated setups such as 32T32R and 64T64R are starting to emerge. The progression towards 5G and its associated features has faced delays in the vRAN/Open RAN space. Achieving parity is crucial as operators work to commercialize features from Release 17 and Release 18, which introduce additional processing complexities and intensify the challenge of maintaining power efficiency.
A recent survey by GSMA Intelligence highlighted energy efficiency as the foremost concern for operators, surpassing even security. This priority stems from both operational and financial necessities and the urgent call for climate action. Many operators are striving to reduce their carbon footprints and attain carbon neutrality, making energy savings in RAN a top priority. GSMA estimates that RAN accounts for an astonishing 73% of operators' overall energy consumption, underscoring the importance of enhancing energy efficiency in base station components.
Section 1.1: Challenges in vRAN/Open RAN
The slow adoption of vRAN and open RAN, despite the long-standing virtualization of core networks, can be attributed to the demanding nature of RAN workloads. The complexity arises primarily in Layer-1 (the physical layer or PHY) processing.
vRAN/Open RAN consists of three main components: the Central Unit (CU), which oversees Radio Resource Control and Packet Data Convergence Protocol functions; the Distributed Unit (DU), responsible for Radio Link Control, Medium Access Control, and PHY; and the Radio Unit (RU), which manages digital-to-analog conversion and MIMO antenna functions.
From a protocol standpoint, the CU handles Layer-3 and part of Layer-2, while the DU manages part of Layer-2 and Layer-1. The RU takes care of the remaining Layer-1 tasks. As one moves from Layer-3 to Layer-1, the complexity and processing requirements increase dramatically. In fact, Layers 2 and 1 together account for nearly 90% of RAN's processing power.
The critical Layer-1 functionalities are divided into Low-Phy and High-Phy components. Low-Phy is managed by the RU, while High-Phy—comprising functions such as demodulation, beamforming, channel coding, and Forward Error Correction (FEC)—is overseen by the DU. These high-demand functions are extremely sensitive to latency and consume a significant portion of the processing power.
Section 1.2: The Role of Compute Architecture
There is a consensus within the industry that dedicated, optimized silicon, like application-specific processors (ASICs), should be used for the RU. The prevailing thought was that general-purpose compute, often referred to as COTS (Commercial Off-The-Shelf), based on x86 or Arm processors, suffices for the CU. However, major data center operators have long recognized that generic compute is inadequate for complex networking and security tasks, such as IPSec and encryption. These responsibilities are usually offloaded to optimized accelerators known as DPUs or Smart NICs.
As the industry embarks on its vRAN journey, it is similarly beginning to acknowledge that COTS servers are also inefficient for High-Phy operations. High-Phy functionalities are crucial to the essence of 5G technology and can significantly impact the performance of vRAN/Open RAN solutions. RAN vendors have invested years in optimizing these functions, which also provide opportunities for differentiation among vendors.
Consequently, the DU will likely consist of a combination of COTS (the host processor) for Layer-2 tasks and one or more accelerators for High-Phy and networking functions. These elements will connect via a standard PCIe interface, commonly used in the IT sector. This arrangement offers numerous advantages: accelerators designed for specific radio workloads are much more energy-efficient, require less cooling, and have a reduced PCB footprint. Scaling becomes easier and more economical, allowing operators to add more accelerators rather than expensive COTS processors to enhance capacity or introduce new features like URLLC.
The first video titled "The benefits of adopting Open RAN architecture" discusses how adopting Open RAN can enhance network flexibility and efficiency.
The second video titled "Deep Dive into O-RAN Architecture" provides an in-depth exploration of the O-RAN architecture and its implications for future network developments.
Closing Thoughts
In conclusion, the ideal compute architecture for vRAN/Open RAN appears to be a combination of dedicated in-line accelerators for High-Phy, ASICs for RU, and COTS host servers for other functions. The challenge for operators lies in selecting the best vendor for their network, focusing on performance metrics such as processing power, capacity, and energy efficiency, along with advanced features like 64T64R massive MIMO and beamforming techniques. Additionally, a vendor's experience in cellular infrastructure is essential.
Ultimately, vendors that excel in these areas will likely thrive in the market. The beauty of open RAN lies in operators' ability to choose the best components—whether COTS, accelerators, or cloud services—creating a true multi-vendor environment. However, this freedom does introduce system integration challenges that warrant further discussion in future articles.
If you're interested in more insights like this and wish to stay updated on mobile and tech industry trends, consider signing up for our monthly newsletter at TantraAnalyst.com/Newsletter or tune into our Tantra's Mantra podcast.