Meta Rolls Out Media over QUIC for Instagram Live Ingest

Meta Ships MoQ for Instagram Live

Meta has begun deploying Media over QUIC (MoQ) for Instagram Live ingest on iOS and Android. The deployment matched the previous RUSH transport across the streaming-quality indicators and reliability measures used for the rollout, and Meta describes it as one of the largest industry usages of the protocol by ingest stream count and total volume.

Live ingest had run on RUSH, Meta's in-house QUIC-based protocol built as a successor to RTMP, for nearly a decade. Shifting to MoQ moves development from a protocol operated primarily by Meta to a shared IETF standards effort, with active participation from Google, Akamai, Cisco, Cloudflare, and academia.

Why a Common Protocol

At Meta, a user who starts a live stream and then makes a voice call in the same app touches three separate protocol stacks, each with its own serving infrastructure and monitoring. MoQ aims to collapse that fragmentation into a single transport for ingest, delivery, and playback.

The protocol's core abstraction is a Track, an addressable stream of media objects pushed by publishers and received by subscribers through relays that cache and forward objects without parsing the payload. QUIC underpins the transport, avoiding TCP-style head-of-line blocking and enabling faster connection setup and connection migration.

MoQ relays provide native fan-out and caching, allow per-subscription latency tuning so interactive participants and buffered services can run at different latencies on the same track, and support track-aware delivery where relays can drop, prioritize, or replay objects based on metadata. When QUIC is blocked, MoQ can fall back to TCP through standardized mappings such as WebTransport over HTTP/2 or the emerging QMUX.

How the Rollout Worked

Meta built a MoQ publisher on moxygen, its open-source MoQ library, integrated with its QUIC stack mvfst and exposed through custom media handlers. A single config parameter selects MoQ or RUSH for a given broadcast.

Rather than rewriting the backend, MoQ ingest connections terminate on dedicated edge endpoints where a translation layer converts MoQ frames to the RUSH format the existing systems already understand. The translator is temporary scaffolding that disappears once the backend consumes MoQ natively.

Server-side routing directs experiment-eligible broadcasts to MoQ-capable endpoints based on experiment-group membership, letting the team resize the rollout in minutes without waiting for app-version adoption. On networks where UDP/443 is blocked or rate-limited, the client picks the transport at the start of a stream based on which handshake wins and uses it for the session.

What Comes Next

Meta plans to extend MoQ across its remaining apps, using MoQ-native constructs like subgroups to carry audio, video, and metadata on independent streams for better resilience under loss. Standardizing on QMUX for TCP fallback would remove the current dependency on RUSH's fallback path and eventually allow RUSH to be retired.

The company is also exploring real-time interactive AI workloads, where a single low-latency, bidirectional transport carrying signaling and media together is a natural fit. The long-term target is an end-to-end MoQ architecture for live delivery and playback, where partial reliability lets players skip stale video after loss instead of accumulating delay, and adaptive bitrate logic trades resolution for continuity as bandwidth changes.

Sources: Engineering At Meta