A Survey of AI Enabled Edge Computing for Future Networks

We have been talking about 5G and beyond (B5G) in IEEE future networks. The Exabyte of data moving from Cloud enabled edge to distributed edge has led to a push for local processing at the edge. We start with what is intelligence and how to distribute it in light of Edge resources. It started with a need for a nimble edge, essentially microdata centers to another extreme "replace every O-RAN and BBUs with a Hyperscale DC". We try to ascertain through this survey can Intelligent devices termed xPU (covering GPUs, FPGAs, SoCs and NVME and many more) provide functional baseline for lowering power consumption & performance, to support AI and ML to deliver edge native low latency & ultra-reliable functions beyond 5G as a better alternative? This is a survey of different industry parties to enable innovation and intelligent edge local processing, discovering challenges and trends to arrive at runtime baseline and abstractions. The survey covers the challenges of different evolving architectures x86, ARM, RISC-V with players like Intel, NVIDIA, Microsoft, IBM, Google & Amazon to name a few. The common theme turns out to be intelligent connected edge services on mini–Data Center as one option while the Edge focus to be Scalable, Secure, Reliable, Performant & Automated platform & Services. We have Reference, Glossary at the end for acronyms used.