At the Cutting “Edge” of AI Capabilities in an Enterprise 5G Network

By Ganesh Shenbagaraman

The topic of edge computing has been around for some time and the inherent advantages of processing at the edge have been well known. With the advent of 5G, edge computing has gained a new impetus. After many false starts and failed experiments that made many realize the challenges in delivering on the promise of edge computing, the answer may very well be the use of 5G technology. So, what is it that makes 5G more suitable for edge computing?

Enhanced Broadband Throughput with Flexibility: With 5G it is possible to deliver several Gbps throughput that can be dynamically adjusted to provide more bandwidth for downloads and uploads as demanded in a dynamic fashion. With Open RAN also gaining momentum, it is possible to choose deployment architectures with distributed and disaggregated RAN and core network elements located flexibly either at multiple locations at the edge or centralized on a cloud.

Ultra-low Latency and Enhanced Reliability: There is a need for very low latency and highly reliable data transfer in use cases like smart factories, healthcare, autonomous driving and other mission-critical applications. With 5G’s URLLC (Ultra Reliable Low Latency Communication) capability, it is possible to achieve latencies of just a few seconds and connection reliability of five 9’s. This is unprecedented and has immense potential for 5G to cater to the most demanding use cases.

Optimized User Plane and Network Slicing: The key to delivering the best Quality of Experience (QoE) to users is in delivering high bandwidth at no or very low latency. 5G RAN and core network architecture is highly optimized for such deployments. In the case of a disaggregated RAN and core network deployment, the RU, DU, CU-UP (user plane) and the UPF can all be located at the edge to serve the users with high throughput while eliminating all latencies. In addition, it is possible to create network slices in an end-to-end fashion to cater to different types of services. The slices allow for differentiation in resource allocation and latency control, in order to meet the service SLAs.

With 3GPP, the standardization body for 5G, focused on enabling edge applications and organizations like 5G-ACIA (https://www.5g-acia.org/) we are seeing a plethora of vendors across the spectrum joining forces to make the 5G-powered edge a reality.

The Rise of Private Networks

With regulators across the globe announcing spectrum policies that encourage industry verticals to build their own private mobile networks, the need for appropriate hardware and software solutions is higher than ever. Industries, enterprises, and campuses are now ready to build networks with CBRS spectrum in the U.S. and with dedicated licensed spectrum for industrial use in many other geographies. Private network solutions must be custom-built and integrated with other applications as per the domain and industry-vertical specific requirements. This allows unique services and applications to be enabled in dedicated networks that provide all the guarantees of data privacy and security.

 

 

 

 

 

 

 

 

NVIDIA and Radisys AI-on-5G Reference Solution

Achieving Ultra-high Performance at the Edge: NVIDIA and Radisys Collaboration

Radisys provides highly scalable 5G RAN and core network components that deliver several Gbps of data at the edge and can be deployed on compact servers. The intense bandwidth demand and low latency processing necessitate high-end processing that is ensured by NVIDIA’s platform. Radisys’ 5G RAN software is integrated with the NVIDIA Aerial PHY implementation. This integrated software leverages low latency processing through inline hardware acceleration using NVIDIA GPUs. Optimized data plane processing is realized using NVIDIA ConnectX-6 DX NICs. This completely packaged software along with NVIDIA libraries for video analytics and AR/VR makes it suitable for edge AI-based 5G deployments. Radisys’ 5G software supports URLLC and network slicing features that can be leveraged by the edge applications.

Edge Computing Revenue (Source: Mobile Experts)

Industry Verticals and Deployments

According to this report by industry analyst firm Mobile Experts, the edge computing market is likely to experience strong growth over the next two decades, with millions of CPUs and GPUs shipped to power hundreds of thousands of edge deployments. We at Radisys forecast that 5G edge AI-based deployments will be used prolifically in industries for large size plants like oil refineries, factory floors, and huge campuses that are consuming and uploading huge video content, sophisticated monitoring and live video/audio analytics using AI/ML capabilities.

Edge is the Future

The future is at the edge in an increasingly data hungry world. It is not only human users that are driving bandwidth demand. It is the millions of devices and sensors that are connected and are constantly transmitting data. All of this demand will be met using high-end computing and processing by state-of-the-art processors: CPUs and GPUs and 5G as a connectivity technology. Welcome to the all-new AI powered edge!

Previous Article
Coming to a screen near you: Radisys at Small Cells World Summit and Private Networks Forum
Coming to a screen near you: Radisys at Small Cells World Summit and Private Networks Forum

Next Article
The Stratosphere’s the Limit: Radisys Joins the HAPS Alliance to Bridge the Digital Divide
The Stratosphere’s the Limit: Radisys Joins the HAPS Alliance to Bridge the Digital Divide