Preview Sessions
summary
Ten-minute rapid tour that names concrete problems, numbers and fixes: NICs hit 120 ns per 100Gb packet so CPU cycles matter; kernel bypass shows a tradeoff between DPDK-style (software-managed) and RDMA-style (NIC-managed) designs with higher CPU cost vs constant NIC state; PIFO-based programmable scheduling and NIC coprocessors become central; three recurring fixes are segmentation, batching and parallelism; previews link Minos/Monoxide/Flashield to realistic workload assumptions and show prototypes and deployment notes.
AI-written summary · about
chapters (12)
- 0:00Host networking basics
- 12:11Distributed systems storage
- 26:05Modern networking hardware
- 37:16Data center analytics
- 49:24Data center architectures
- 1:06:48Wireless communication
- 1:22:56Network modeling
- 1:39:39Operating systems
- 1:53:50Network characterization
- 2:07:15Monitoring and diagnosis
- 2:19:55Machine learning systems
- 2:30:10Network functions
related talks
01h 24mPreview Session
12mPreview: Networking
25mDPDK performance optimization for edge computing scenario
27mLIVE Q&A #4
20mCan OVS-DPDK be further optimised?
14mUnderstanding RDMA Microarchitecture Resources for Performance Isolation
17mPerformance Interfaces for Network Functions
31mA Practical, Research‑Backed Introduction To Cloud‑Native Kernel‑Bypass Networking