This blog is co-authored by Mike Moore, Telecom Solutions Marketing Consultant, Dell Technologies.
If there is one thing to recognize about 5G that, above all the hype, 5G has ignited Enterprise’s edge transformation plans. It looks the market with its speed records and shallow latency figures. But it was also able to convince critical business operations of its security and reliability attributes. Now, every company we talk to is working on incorporating process automation and using data for fast and consistent decision-making.
Jack of All Trades – Master of Them All
5G is a big step forward in all dimensions (speed, latency, number of users, security, available spectrum, flexibility), and has the capability to become the target architecture for the years to come. Expecting consolidation gains, many CSPs will converge legacy networks and repurpose spectrum assets in favor of 5G.
5G’s success results from the wider spectrum availability and the intelligent way 3GPP planned its releases and designed its flexible logical Tx/RX frames, allowing it to independently scale capacity, throughput, and latency. This flexibility allows it to adapt and deliver different connectivity services, always at a very effective cost.
5G’s also benefits from more advanced modulation techniques, modern electronics and its rumble approach to “Openness” – 3GPP created 5G to be open and leverage contemporary innovation streams such as Edge computing, and open IP protocols, CUPs, and open APIs. Proof of this are 5G’s expected longevity and sequential releases where 3GPP keeps adding new functionalities to enable it to serve different use cases.
AI for 5G and 5G for AI
Artificial intelligence is intrinsically embedded in many of 5G’s signal processing tasks, traffic prediction algorithms and self-optimization routines. With the use of AI, 5G networks can predict traffic patterns and electronically focus its antenna array accordingly, assuring that network resources will always be used effectively. It can also intelligently power off parts of the network equipment to save energy.
Advanced scheduling mechanisms and noise cancellation processing tasks are also 5G processes that rely on AI. By combining big data, IoT and AI, disruptive technology advancements start to revolutionize traditional industry verticals.
At the time CSPs and hyperscalers deploy the edge network to implement 5G networks, by nature, they will be concomitantly enabling the necessary computing infrastructure to host AI workloads. The available edge cloud resources will serve different players, with abundant computing and connectivity resources at the edge. Innovative business models will drive from this flexible resource assignation generating a matrix of innovative cooperation models.
At Dell Technologies, we see the combination of 5G, AI and data connectivity at the edge as a transformative platform to enable new possibilities for enterprises, governments and society in general. AI will allow machines and systems to function with intelligence levels like that of humans. 5G will connect sensors and devices at speed while AI simultaneously analyzes and learns from data, enabling real-time feedback loops.
Trends in GPU/FPGA/Other Edge AI Acceleration Approaches
When it comes to edge AI Acceleration options, there are several distinct options available. This allows for a tiering of acceleration capabilities, fitted to the demands of the AI/ML application while taking into consideration other factors like cost, space, power consumption and heat dissipation.
Additional AI/ML capabilities are also being integrated into upcoming CPU Architectures of various chipset vendors. As an example, Intel’s upcoming 4th Generation Xeon Scalable Processor will include a new instruction set for deep learning performance, with matrix multiplication instructions that promise to deliver much improved AI/ML performance available by default. This will enable the movement of AI/ML functions to the edge, without the need for add-on accelerators, such as GPUs or FPGAs, and do so in a power and space-efficient manner.
For increased AI/ML processing at the Edge, though, PCIe Accelerator Cards are the way to go. As an example, Dell’s edge/Telecom tailored PowerEdge XR11 and XR12 are ruggedized, short depth, NEBS Level 3 and MIL-STD certified servers, supporting the expansion of AI/ML capabilities via PCIe, providing a tiering of acceleration options for the edge.
3GPP Standardizes 5G AI-based Procedures
3GPP wants to 5G becomes a relevant part of this AI/ML innovation fabric at the edge, and seamlessly integrates into the broader edge cloud in a more extensive manner.
Via a Technical Report (TR), TR 22.874, Technical Specification, 28.105, and the new TS 28.908, 3GPP will provide a standard approach to the complete lifecycle management of AI/ML Enabled Functions in the 5G network.
Complex network capabilities will be offloaded to AI/ML models to leverage the ability to constantly adapt and optimize for the latest network conditions. Network Planning, Management, and Performance Optimization (including SON) shows great promise of increasing network performance and reliability while reducing the overall costs of network administration.
Also, by using open northbound open API interfaces, 3GPP will enable the formation of an ecosystem of SW developers working with supported development tools extensible for AI and ML functions.
Technology for Good
A new data-aware society is forming at the edge, with 5G and AI capabilities excelling not only productivity levels but also long aimed at standards and environmental sustainability.
Thoughtful applied technology has the power to improve operational processes in retail, manufacturing, banking, transportation, healthcare and government. It will reshape modern society’s expectations and possibilities, allowing us to dream about a better world and help address some of humankind’s more challenging issues. The power of technology to transform our lives has never been so realistically reachable, but it’s still a race against time.