How NetEase Cloud’s QUIC Optimizations Boost Audio‑Video Calls in Weak Networks

This article explains how NetEase Cloud Communication leverages the QUIC protocol—detailing its transport‑layer advantages over TCP, protocol‑level differences, and specific optimization practices such as multiplexing, unreliable DATAGRAM frames, compression, and dynamic redundancy—to dramatically improve audio‑video signaling latency, loss tolerance, and bandwidth utilization in challenging network conditions.

NetEase Smart Enterprise Tech+
NetEase Smart Enterprise Tech+
NetEase Smart Enterprise Tech+
How NetEase Cloud’s QUIC Optimizations Boost Audio‑Video Calls in Weak Networks

Introduction

NetEase Cloud Communication, a leader in audio‑video services, focuses on delivering reliable calls even under extremely weak network conditions. The article outlines the company's practical optimizations of the QUIC protocol.

QUIC Advantages Over TCP (Transport Layer)

0‑RTT Connection : Built on UDP, QUIC eliminates handshake latency, using TLS 1.3 early_data to transmit encrypted data after a single round‑trip.

Multiplexing / No Head‑of‑Line Blocking : Unlike HTTP/2, QUIC streams are independent, avoiding head‑of‑line blocking.

Connection Migration : QUIC uses a 64‑bit ConnectionID, allowing seamless migration when the four‑tuple changes (e.g., Wi‑Fi to cellular).

Customizable Congestion Control : Congestion control algorithms are implemented in the application layer, enabling developers to iterate and optimize.

Protocol‑Level Differences Between QUIC and TCP

Separate Packet Number Spaces : Four encryption levels each have distinct packet number spaces.

Monotonically Increasing Packet Numbers : Guarantees unambiguous retransmission handling; data order is indicated by stream offsets.

Clearer Loss Epoch : Loss detection periods are refreshed on each acknowledged packet, allowing more precise congestion‑window updates.

No Reneging : Once a packet is acknowledged, it cannot be declared lost, simplifying protocol design and reducing memory pressure.

More ACK Ranges : QUIC ACK frames can acknowledge many more ranges than TCP’s SACK, speeding recovery in high‑loss scenarios.

Explicit Delayed ACK Correction : ACK frames embed measured acknowledgment delay for more accurate RTT estimation.

Probe Timeout (PTO) Replaces RTO/TLP : PTO incorporates the peer’s expected maximum acknowledgment delay and does not collapse the congestion window on timeout, allowing aggressive sending.

Minimum Congestion Window of Two Packets : Improves resilience compared to TCP’s single‑packet minimum.

QUIC in NetEase Cloud’s Audio‑Video Service

Signaling (SDP exchange, room creation, user info) traditionally uses WebSocket over TCP, which suffers from high connection latency and weak‑network performance. NetEase replaces WebSocket with QUIC and applies several optimizations:

Multiplexing : High‑priority STREAM for request/response messages; lower‑priority STREAM for heartbeat.

Unreliable Transmission (DATAGRAM) : Notify messages that require low latency but not reliability are sent via QUIC DATAGRAM frames, declared in the Initial packet’s transport parameters.

Compression : Deflate algorithm compresses STREAM frames to reduce bandwidth.

Dynamic Redundancy : Redundancy level adapts to RTT and loss rate, enhancing weak‑network robustness.

Performance in Weak Networks

First‑Frame and Login Latency : QUIC reduces first‑frame latency by ~20% and login latency by ~30% compared to TCP, mainly due to 2–3 RTT handshake savings.

Loss Tolerance : QUIC maintains service up to 70% upstream and 75% downstream packet loss, a 50% improvement over TCP, thanks to dynamic redundancy and superior congestion control.

Bandwidth Utilization : Under bandwidth constraints, QUIC consistently achieves >90% utilization, far surpassing TCP.

Outlook & Summary

NetEase Cloud has significantly improved reliable data transmission with QUIC, yet challenges remain, such as reducing unnecessary redundancy when only one direction experiences loss and further optimizing congestion‑control algorithms for high‑RTT, high‑loss environments.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

backendQUICTransport ProtocolAudio-Video
NetEase Smart Enterprise Tech+
Written by

NetEase Smart Enterprise Tech+

Get cutting-edge insights from NetEase's CTO, access the most valuable tech knowledge, and learn NetEase's latest best practices. NetEase Smart Enterprise Tech+ helps you grow from a thinker into a tech expert.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.