From the course: AI in RAN (Radio Access Network): Transforming Mobile Networks
Unlock this course with a free trial
Join today to access over 25,300 courses taught by industry experts.
The AI-RAN network challenge
From the course: AI in RAN (Radio Access Network): Transforming Mobile Networks
The AI-RAN network challenge
Imagine a superhighway that must handle two different types of traffic. One lane requires vehicles to pass at exactly timed intervals and the other lane must handle the unpredictable massive truck convoys that all need to arrive at the same time. Running 5G radio access network and AI workloads on the same network is this exact challenge. In this video, we'll explain the unique technical demands that make generic network equipment insufficient for AI RAN. We'll look into the key functional requirements that drive the need for specialized switches and the network cards. The need stems from a fundamental conflict in network traffic patterns. AI workloads, especially the training, generate a massive synchronized data between hundreds or thousands of processors which require immense bandwidth. On the other side, for 5G RAN, it has a non-negotiable requirement for ultra low latency. A standard data center network optimized for North-South client server traffic is not built to handle either…
Contents
-
-
-
-
-
-
-
(Locked)
The AI-RAN network challenge2m 56s
-
(Locked)
Programmability in RAN through SDN principles3m 34s
-
(Locked)
Overview of software-defined networking (SDN)2m 46s
-
(Locked)
Deploying software-defined networking (SDN)5m 8s
-
(Locked)
Multi-access edge computing: The key to ultra-low latency4m 54s
-
(Locked)
MEC deployment: Guide to edge orchestration using Open RAN4m 14s
-
(Locked)
-
-
-