In this series focused on the benefits of a disaggregated Open RAN, we have had a chance to see how an open RAN has helped expand the ecosystem and led operators to consider new vendor solutions that meet their needs. Standards have emerged which help ensure increased interoperability. In the last blog post we covered system integration and pre-deployment aspects of disaggregated RAN solutions in detail. In this latest entry to our series, we’ll take a look at how disaggregated solutions compare in terms of benchmarking performance.
Performance and deployment readiness
There are a couple of questions that are often brought up: Will an open disaggregated multi-vendor RAN solution perform as good as the traditional (closed) RAN solutions? Do these disaggregated components. as a whole meet the various needs of an operator deployment?
The answers to these questions lie in the way open and disaggregated solutions are built. These are of course vast and complex topics. We must take into account the hardware, software and other aspects critical to building a high-performance RAN solution. Let us look at some of the key ingredients of such a solution. The picture below illustrates these aspects.
For Open RAN solutions to prove that they match up head-to-head with traditional solutions, we need to understand how commercial off-the-shelf (COTS) products can come together to form a complete solution.
The hardware question
RAN products demand high performance hardware, including the processors that can perform compute intensive tasks while being power optimal and specialized hardware for certain functions and other aspects like cooling, power consumption and environmental readiness.
Processors – At its most basic level, performance of the products depend on the processors to a great extent. Like most traditional solutions, Open RAN solutions are driven by x86 or Arm-based general purpose processors (GPPs). By choosing appropriate GPP based servers, operators can easily find solutions which match traditional solutions and meet their performance needs. The processor companies are also addressing the Open RAN market requirements by releasing processor variants with extra smarts needed for some of the intensive compute operations typical of a RAN product.
Hardware Acceleration – Open RAN solutions are increasing performance through hardware accelerators like field-programmable gate arrays (FPGAs) and application-specific integrated circuits (ASICs). While targeted for specific, custom functions, these hardware accelerators are commercial off-the-shelf products and are widely used by vendors, including traditional ones. Selecting the right chipsets and integrating them correctly is key to achieving top performance.
Other Hardware Aspects – Power consumption and cooling are hardware considerations operators must take into account. The power consumption of a solution is an important metric because the correct power must be ensured at the cell sites. There are limits to how much power can be made available and hence a proper understanding of a solution’s power consumption profile is essential. Cooling becomes an important aspect as processors have different cooling needs; choosing one processor over another may have an impact on cooling. Additionally, the environmental conditions for the deployment, the high and low temperature range, and its impact on the enclosure, must also be considered when selecting a server package for deployment.
Low latency OS and network processing
After the hardware considerations are evaluated, there are performance factors to be considered stemming from the software side of the solution.
Real-time pperating system – There are many standard versions of the Linux operating system (OS) available for deployment. OSBy using real-time Linux flavors hardened for RAN/edge deployment, and you can achieve low latency in the operating systems. These real-time Linux OSs have been tuned to perform better than general Linux, allowing Open RAN solutions to perform in-line with those from traditional vendors. Even in cloudified RAN implementations, such high-performance OSs are needed to obtain high performance.
Fast data path – The rate at which data comes into the RAN and goes back out to the network is an essential part of measuring performance. There have been tremendous gains in performance and high throughput of packets while not consuming so many core processors. Specifically, use of Data Plane Development Kit (DPDK) to enable fast packet processing on x86 platforms, and Vector Packet Processing (VPP) algorithms, allow Open RAN solutions to pass the packets in the radio network from the handset to the base station and to the core network faster. The faster the packets are transferred, the lower the latency.
Software performance
As RAN software has been decoupled from the hardware, it is important to understand how it performs in different deployment scenarios. There are standard base station benchmarks on performance and capacity to objectively measure a RAN product. These include capacity aspects like number of connected users, scalability of base station components to support multiple cells, carriers and radios. The signaling performance is measures by latencies in end-to-end messages, number of messages handled per second etc. The most important set of metrics are the ones of total throughput, number of users scheduled for data traffic every second, CPU utilization etc. Building such high-performance software that runs on COTS hardware and open source environments and still handles all the complexity of 3GPP protocols is a challenge. And by just delivering that Open RAN solutions are making headway today across many operator networks.
Operational aspects and field performance
Now that the Open RAN hardware and software’s performance have been assessed, we need to look at the final areas that must be considered in any performance discussion, specifically its deployment and operation in the field.
Deploying Open RAN solutions – With COTS hardware coming in to play, there is more “IT” flavor to the RAN equipment which is a change for those who have deployed traditional vendor specialized hardware-based equipment. And the aspect of cloudified RAN implementations is another new dimension. Deploying an Open RAN solution, with open interfaces and a more modern approach to management interfaces, is a different experience for service providers’ operations teams . As the industry is evolving and readying for a mainstream Open RAN experience, more and more operators are educating and equipping their personnel to handle Open RAN architecture. The providers of Open RAN solutions too are working with operators closely in this transition journey .
Operations and maintenance – A smooth roll out of Open RAN solutions is dependent on the ease of configuring, monitoring and maintaining RAN network nodes. By adopting modern element management and integrating with sophisticated orchestrators, Open RAN solution providers are improving efficiency and automation in managing RAN networks while delivering OpEx benefits.
Field performance vs. lab performance – Even after elaborate testing at an operator’s labs, there could be gaps seen in the way the products perform in real-life conditions. This is true of any RAN solution, open or otherwise. The most effective way to deal with this is for the operators doing extensive field trials to identify such gaps and for the vendors to work closely with the operators to address any gaps. It is quite normal for operators to work closely with vendors, several months post deployment to monitor and improve network performance. The good news is that in such open solutions, a lot of the metrics, measurements and interface details are all done in an open and transparent manner. It is easy to spot and address improvement areas on a continuous basis, the real goal of having such open solutions deployed!
We addressed some key aspects of performance and deployment readiness of Open RAN solutions and how to get it right. Most operators are now actively engaged with vendors in selecting and evaluating Open RAN solutions. These operators are seeking to understand the benefits and improvement areas through extensive trials. The learning experiences from various operators are brought into multiple industry forums creating opportunities for appropriate standards and ecosystem offerings to improve.
This cycle is destined to bring in innumerable gains, big and small, to all who partake in the Open RAN movement. In such an environment, is there an area where there is maximum amount of interest from vendors and operators? The answer is yes, the Cloud RAN solution. We will take a deep look at that in our next blog.
