How Netflix Scaled Live Streaming Ops to 400+ Events a Year

Rough Beginnings

In March 2023 the engineer who built Netflix's first live‑streaming pipeline also performed all operational duties. There was no dedicated ops team or command center; incident‑response runbooks were written for SVOD, and SLAs ignored live‑stream latency requirements. Early shows were monitored from a laptop via Slack while engineers manually troubleshooted under millions of viewers.

Physical setup was equally ad‑hoc: a makeshift control room in a conference room, rented broadcast hardware for large events, and a multi‑camera multiviewer. Every live event required intensive collaboration across engineers and leadership.

Broadcast Operations Architecture

The core challenge for a tech company entering live broadcast is merging traditional TV practices with large‑scale streaming engineering, centered on the Broadcast Operations Center (BOC). The BOC receives full‑production video from venues, processes it (including captioning, graphics, ad insertion), and forwards it to the streaming pipeline using a hub‑and‑spoke model and redundant links such as dual internet circuits and SMPTE 2022‑7 seamless switching.

To guarantee signal reliability, Netflix mandates three fully independent transmission paths for any primary stream, preferring dedicated fiber and satellite feeds, then enterprise‑grade internet, then Secure Reliable Transport (SRT). On‑site hardware is fully redundant with dual power supplies, UPS, and surge protection. Before each live event, a detailed “FACS/FAX” facility‑check runs audio‑video sync, latency, and quality tests.

Evolution of Operational Models

Phase 1 – “All‑Engineers” : Engineers who built the pipeline also manually operated each broadcast, a non‑scalable model.

Phase 2 – Specialized Engineering : Creation of Streaming Operations Engineering (SOE) to configure events and handle first‑line issues, freeing core developers. As physical broadcast grew, Broadcast Operations Engineers (BOE) were added to own hardware and facility workflows.

Phase 3 – “Co‑Pilot” Control Room : Two Broadcast Control Operators (BCO) work as pilot and co‑pilot, suitable for 1–2 daily events but insufficient for a ten‑event‑per‑day schedule.

Phase 4 – Transmission Operations Center (TOC) Fleet Model : All events are managed as a fleet with three tiered roles:

Transmission Control Operator (TCO) – manages inbound feeds (fiber, SRT, satellite), handling up to five events simultaneously.

Streaming Control Operator (SCO) – manages outbound feeds, also up to five events.

Broadcast Control Operator (BCO) – focuses on creative execution, maintaining a strict 1:1 operator‑to‑event ratio and handling subtitles, SCTE ad‑insertion metadata.

For flagship “Big Bet” events (e.g., major NFL games) a dedicated “Big Bet Model” provides a full‑time BOC, exclusive engineers, and premium equipment to guarantee the highest reliability.

Live Command Center (LCC)

The LCC is not a traditional Master Control Room or NOC; it provides end‑to‑end visibility of every live stream, from venue capture through cloud encoding, CDN delivery, and playback. It runs a dedicated observability stack that aggregates telemetry such as concurrent viewers, start‑failure rate, re‑buffer ratio, CDN health, encoder status, and signal‑path health.

During a live event the system processes up to 38 million events per second. Two roles run the LCC:

LCC Operations Leads – shift supervisors and incident commanders who triage alerts, decide escalations, and drive the response workflow.

Live Technical Launch Managers (TLMs) – act like air‑traffic controllers, coordinating over 45 cross‑functional teams months in advance to ensure runbooks and escalation paths are ready for 2 am incidents.

Events are tiered into Low‑Profile, High‑Profile, and Big Bet, each with a corresponding Live Operational Level (LOL) for non‑ops teams:

Red – online throughout the event (e.g., major boxing).

Orange – online 30 minutes before start, then downgraded after first ad.

Yellow – off‑line but must respond within two minutes via PagerDuty.

Grey – business‑as‑usual, contacted only via normal paging.

Operational Principles and Culture

Reliability precedes efficiency: standardized runbooks, tiered incident structures, and pre‑recorded failure modes ensure that the 50th show runs as smoothly as the 5th. Real‑time monitoring tools with sub‑second latency are essential; the Live Control Center was built as a product decision to turn millions of telemetry events into actionable insights for a small ops team.

Documentation and onboarding are critical. Detailed runbooks enable new contract operators to become fully competent within a week, turning once‑exceptional events into routine operations. The “squad model” limits communication channels and provides a single escalation path to the LCC, reducing chaos during massive events like the Tyson‑Paul fight, which saw over 300 online participants and 64 million concurrent streams.

Future Outlook

By early 2026 Netflix plans new Live Broadcast Operations Centers in Los Angeles and a Live Operations Center (LOC) in London, establishing a follow‑the‑sun coverage model. Expected 2026 live volume exceeds 400 events, including a 24/7 free linear channel with TF1. Continued automation of alerting and exception‑based monitoring aims to further lower manual workload.