Our live-streaming redundancy stack — what fails, and what we fail to

Recorded work has a safety net called 'tomorrow'. Live has none. Everything we design for a stream assumes each single point will fail at the worst possible moment — so nothing important is ever a single point.

Two ways out of the building

Uplink is where most streams die. We bond a hard line with bonded cellular so the encoder keeps a path out even if one drops entirely. If the venue's network is the only option, we treat it as untrusted and bring our own.

A hot spare that's already running

A backup encoder that needs switching on is not a backup. Ours runs in parallel, already encoding, so failover is a cut rather than a reboot. The switch is rehearsed before doors open, not discovered during the keynote.

Record locally, always

Every camera records in-body while it streams. If the line degrades, the show still exists in full quality on cards, and a clean recap or VOD is ready regardless of what the internet did that afternoon.

Rehearse the failure, not just the run of show

We pull a cable on purpose during tech rehearsal and time how long recovery takes. If the answer is 'we're not sure', it isn't redundancy yet — it's a hope.

• Bonded hard line + cellular uplink, venue network never trusted alone
• Hot-spare encoder running in parallel, failover rehearsed
• In-camera recording on every source as the last resort master
• Branded holding slate ready to cut to in one button
• A named person watching bitrate, not the show