Meta deployed Media over QUIC (MoQ) for Instagram Live ingest, transitioning from their proprietary RUSH protocol. MoQ, an emerging IETF standard built on QUIC, aims to unify media transport for various use cases like live streaming, real-time communication, and AI, improving latency, resilience, and operational efficiency. This shift enables a more streamlined architecture by moving to a single, open-standard protocol across the entire media lifecycle, while also facilitating adaptive bitrate logic and native fan-out capabilities.
Read original on Meta EngineeringHistorically, various media experiences (live streaming, voice calls) within Meta relied on disparate protocol stacks like RTMP, HLS/DASH, WebRTC, and RUSH. This led to operational complexity, with separate serving infrastructures, monitoring, and operational expertise required for each. As live video, real-time AI, and generative AI converge, the need for a unified, low-latency, and resilient media transport became critical to reduce maintenance overhead and foster innovation across product surfaces.
MoQ is an open IETF standard designed for low-latency, scalable media delivery over the QUIC transport layer. It acts as a foundational protocol intended to consolidate ingest, delivery, and playback across different use cases. Meta, having previously developed RUSH (a QUIC-based protocol for live streaming), contributed its production experience to the development of MoQ, making it a natural evolution for their media infrastructure.
Key Properties of MoQ
MoQ leverages QUIC's benefits like head-of-line blocking avoidance, faster connection setup, and migration. It inherently supports fan-out and caching via 'relays' which efficiently scale media streams. Tunable latency per subscriber, track-aware delivery for better congestion control, and first-class media context for adaptive bitrate are core to its design. Importantly, MoQ includes a standardized TCP fallback, ensuring resilience on networks where QUIC might be blocked.
Meta's initial deployment targeted Instagram Live ingest due to its scale and the desire to move from proprietary RUSH to an open standard. The rollout involved integrating a MoQ publisher (built on Meta's open-source moxygen library) into client apps, and a crucial translation layer at the edge to convert MoQ frames to the legacy RUSH format, allowing the existing backend pipeline to remain unchanged during the transition. Server-side routing facilitated phased rollouts independent of app version adoption.
The successful rollout to Instagram Live ingest achieved metric parity with the previous RUSH system, validating MoQ's performance at Meta's scale. This sets the stage for extending MoQ across the entire end-to-end live video architecture, enabling lower latency, better resilience, and richer interactive experiences, and potentially unifying other real-time interactive AI workloads under a single transport protocol.