Networking @Scale Summer 2022
as the organisers put it
Top-level summary: Building and operating large-scale networks hosting applications that serve billions of people worldwide often present complex engineering challenges to solve. At the recently held Networking@Scale 2022 virtual conference hosted by Meta on Jun 01 and Jun 02, 2022, engineers from Cloudflare, Fastly, Google, Microsoft Azure, Netflix, and Meta presented talks and engaged in […]
who it's for
transport protocol engineers, network engineers, platform engineers, site reliability engineers, kernel engineers, CDN engineers, infrastructure engineers
topics
QUIC zero-RTT handshakes; QUIC congestion controller tuning; QUIC cache direct server response (DSR); QUIC jumpstart connection seeding; HTTP/3 interoperability and diagnostics; DCTCP with ECN for datacenter congestion; eBPF-based fleet network tuning (NetEdit); sockops and scavenger congestion control agents; qlog production tracing for QUIC/DSR debugging; NATless IPv6/IPv4 syscall interception (addfd swap); NIC offload for encryption and pacing
| title | author | length |
|---|---|---|
| Opening Remarks | Omar Baldonado | 06m |
Conference frames transport as the concrete problem of moving data during the pandemic driven surge, citing published papers on pandemic surge and network… | ||
| The Future With QUIC | Jana Iyengar | 19m |
QUIC ships TLS inside an encrypted, UDP-based transport to avoid ossification and enable rapid user-space deployment; it delivers zero-RTT transport handshake… | ||
| Quick Cache DSR | Matt Joras, Yair Gottdenker | 13m |
Implements a QUIC cache DSR where the front-end proxy writes QUIC headers and RPCs compact send instructions to a back-end cache that completes and sends… | ||
| Improving Transfer Times in the Backbone Network Using QUIC Jump Start | Joseph Beshay | 18m |
Meta deploys QUIC jumpstart with 0-RTT and a modified congestion controller that keeps a per-process EWMA of cwnd and min RTT per region, sets a pacing rate… | ||
| LIVE Q&A #1 | Bharat Parekh | 25m |
Multiple QUIC versions already coexist so kernel implementations do not kill agility but slow rollout of specific versions; interoperability relies on spec… | ||
| Layer Four and Three Quarters: Fantastic Quirks and Where to Find Them | Lucas Pardue | 31m |
Cloudflare Protocols team examines real-world quirks between layer 4 and 7 with concrete findings: traffic is overwhelmingly encrypted (about 99% HTTPS)… | ||
| The Challenges of 0-RTT in IETF QUIC | Ian Swett | 16m |
IETF QUIC 0-RTT causes real failures and slowdowns: servers hit a 3x amplification limit, multiple packet number spaces hide which keys peers have, and PTO… | ||
| LIVE Q&A #2 | Luca Niccolini | 23m |
Live Q&A focuses on practical QUIC and HTTP/3 issues: adoption grows via browser defaults and DNS service records but many systems stay on HTTP/1.1 for… | ||
| Tackling DC Congestion and Bursts | Balasubramanian Madhavan, Abhishek Dhamija | 19m |
Deploys DCTCP with ECN for intra-region flows using a custom cgroup BPF selector, fixes a kernel SYN-ECN bug with a temporary BPF marker, and measures adoption… | ||
| NetEdit: Fine-grained Network Tuning at Scale | Prashanth Kannan, Prankur Gupta | 19m |
NetEdit is a BPF-based agent deployed fleetwide on 1M+ servers to apply per-service and per-connection kernel tweaks without changing the kernel. It exposes a… | ||
| LIVE Q&A #3 | Neil Spring | 22m |
Team uses eBPF to tune transport in-kernel without rewriting TCP, packaging programs as a single binary golden image for fast rollout and minute-scale… | ||
| NATless IPv6/IPv4 Address Translation | Keerti Lakshminarayan, Alok Tiagi | 18m |
Demonstrates a kernel‑level syscall interception that makes IPv6-only containers reach IPv4 services without per‑packet NAT by intercepting connect, switching… | ||
| Network Entitlement: From Hose-based Approval to Host-based Admission | Guanqing Yan, Manikandan Somasundaram | 17m |
Meta moves from pipe-based contracts to a host model that replaces O(n^2) pipe agreements with O(n) host entitlements, uses a tree‑based regression (historical… | ||
| RDMA @Scale | Jitu Padhye | 25m |
Azure moves most storage traffic off CPUs: about 65% of data-plane traffic uses RDMA, across links up to 100 km and what the talk claims as an exabyte-scale… | ||
| LIVE Q&A #4 | Ying Zhang | 27m |
Panel covers concrete designs and operational numbers: a host-side TSA intercepts connect calls, swaps file descriptors across namespaces using seccomp to… | ||
| nothing matches | ||