Most AI Pilots Fail to Scale. MIT Sloan Teaches You Why — and How to Fix It
The Most Addictive Python and SQL Courses
Overview
Google, IBM & Meta Certificates — All 10,000+ Courses at 40% Off
One annual plan covers every course and certificate on Coursera. 40% off for a limited time.
Get Full Access
In this engaging 21-minute keynote, Professor Stefan Schmid from TU Berlin explores how emerging optical switching technologies are transforming datacenter networks from fixed architectures to dynamic, demand-aware systems. Discover how Reconfigurable Datacenter Networks (RDCNs) can dramatically improve efficiency by adapting to traffic patterns in real-time—similar to changing highway lanes based on congestion. Learn about the technological enablers like tunable lasers and programmable mirrors that make this possible, and understand the complexity map showing how different workloads (batch processing, deep learning) exhibit varying spatial and temporal structures. Explore the limitations of traditional fixed topologies, the benefits of reconfigurable architectures, and the challenges in routing and congestion control. Drawing from his Communications of the ACM article and supported by real-world examples like Google's Jupiter network, this forward-looking talk provides both high-level vision and technical depth on how dynamic topologies can significantly enhance bandwidth utilization and energy efficiency in modern datacenters.
Syllabus
Revolutionizing Datacenter Networks via Reconfigurable Topologies | Stefan Schmid | TU Berlin
Taught by
SAIConference