Learn Generative AI, Prompt Engineering, and LLMs for Free
Lead AI-Native Products with Microsoft's Agentic AI Program
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
Learn about addressing performance bottlenecks in RDMA-based container networks through this 15-minute conference presentation from NSDI '25. Explore how RDMA-offloaded container networks (RCNs) face unexpected performance degradation when scaling to millions of containers in data centers, with researchers identifying RDMA NICs (RNICs) as the primary source of scalability walls. Discover the innovative approach of using combinatorial causal testing to infer RNIC architecture models and performance characteristics despite limited visibility into hardware internals. Examine the ScalaCN system design that proactively optimizes network function offloading schedules, achieving 1.4× improvement in end-to-end network bandwidth and 31% reduction in packet forwarding latency while resolving 82% of identified performance causes. Understand how this research methodology successfully identified and reported RNIC performance issues to vendors, leading to confirmed fixes and ongoing collaboration for hardware improvements in large-scale container networking environments.
Syllabus
NSDI '25 - Mitigating Scalability Walls of RDMA-based Container Networks
Taught by
USENIX