Mature Applications of Real-Time Audio/Video in Education: TalRTC Architecture, High Availability, and Network Optimization

Overview

On June 9, 2021, the Asia‑Pacific Content Distribution Conference & CDN Summit in Beijing featured a talk titled “Mature Applications of Real‑Time Audio/Video in Education” where Feng Quancheng introduced GoodFuture’s achievements in RTC technology and its proven use cases in education.

1. Live Streaming Platform Overview

The platform centers on the TalRTC streaming server and provides a full‑scene product matrix, including a native UDP‑based RTC SDK, a standard WebRTC Web SDK, and IM/signaling integration to enable real‑time classroom communication. The server can transcode RTC streams to RTMP for CDN distribution, supporting millions of concurrent viewers, and includes a VOD cloud for recording, storage, editing, stitching, and transcoding.

The product is divided into four blocks: RTC website & developer center, InnerAPI & backend management, RTC server & SDK, and the service‑quality assurance system (cockpit management, alarm system, log collection, and big‑data analysis).

2. TalRTC Overall Architecture

The architecture consists of three layers:

a) Client SDK Access – Clients resolve DNS to an SLB load‑balancer, which assigns an IPLocation server for optimal media node scheduling.

b) Business Servers – Includes IPLocation (scheduling), RoomManage (room state), RECMGR (recording management), and HeartServer (heartbeat and Redis‑based resource reporting).

c) Media Servers – Comprises SignalGW (signaling), Audio (audio forwarding), AudioMix (audio mixing), Video (video forwarding), Record (recording), and LocalLog (log persistence).

3. High Availability and Weak‑Network Strategies

RTN scheduling employs regional, carrier, and real‑time analytics to select primary and backup media nodes; clients automatically switch to backups on failure. On the SDK side, SVC Controller provides layered encoding (size‑based and temporal scalability), while server‑side SVC Filter adapts streams to network conditions. Additional mechanisms include Sender‑Side BWE, REMB, NACK/FEC for packet loss recovery, and Pacer for smooth packet emission.

4. Special Optimizations for Education Scenarios

Since 2018 GoodFuture has developed custom NACK/FEC algorithms and Pacer strategies to reduce stutter in weak networks. AI‑enhanced audio 3A technologies (AI ANS for non‑stationary noise reduction and AI AEC for nonlinear echo cancellation) are integrated via collaboration with the AI research institute. Stream forwarding prioritizes teacher streams, and a fallback degradation strategy (video+audio → video‑only → audio‑only) ensures students always receive at least the teacher’s audio under poor network conditions.

The TalRTC solution received the “RTC Technology Innovation Award,” recognizing its technical excellence and mature application in education.