LED playout redundancy The main PC dies at 20:41. Nobody in the stand notices.
Every club asks this in the first call, and most software answers it with the word "redundancy" and no detail. Here is exactly what happens, in the order it happens, and what it needs from your rack.
Four numbers, and the gap between two of them.
A second machine that follows the match.
A spare PC with the software installed is not redundancy. Somebody has to notice the wall has gone, find the machine, open the project, work out where the rotation had got to and start it, and by then the half is over and the sponsor has lost its seconds.
A USM backup PC runs the whole match alongside the main one. It holds the same project, it follows the same cues, and it knows which slot is on the wall right now, because it has been following all along. What it does not do is drive the wall, until the moment it has to.
The two machines talk on the venue network once a second, and that is the failure detector. There is no cloud service, dashboard or person in the loop. If the main stops answering, the backup already has everything it needs to take over, and it does.
What a dead PC looks like from the stand.
The times are the intervals the software works to, laid on a clock so the gap is visible.
- 20:41:02
The main PC stops
Power, a kernel fault, a pulled cable. From the wall’s point of view the picture is still there: the LED processor is holding the last frame it was sent.
- 20:41:03
The first heartbeat is missed
The backup was expecting a beat and did not get one. One missed beat is not a failure; a machine can be busy for a second.
- 20:41:07
The backup takes the wall
Five seconds of silence is the threshold in the code. The backup promotes itself and drives the wall. It has been mirroring the master’s cues all match, so it knows the playlist and where the match had got to.
- Half-time
Somebody deals with the main PC
Handing the wall back is a decision somebody makes, never automatic. Two machines both believing they are in charge is the one failure worse than the first, so USM makes an operator say which is which, and warns loudly on both screens if it ever sees two.
What we would ask you about your rack.
- Both machines need to see the processors. The backup drives the same LED processors as the main one. If only one PC is cabled to the wall, the second one cannot take over.
- Several screens stay together. A ground with a perimeter run and a fascia does not fail over one screen at a time. The mirroring is by position, so screen two on the backup is screen two on the main.
- The project is the same project. Both machines take the same project from the cloud, so a change made in the office an hour before kick-off is on both. A backup running last week’s sponsors is worse than no backup.
- Put them on different power. Two PCs on the same extension lead fail together. UEFA asks for dual signal servers each on a UPS for its own competitions, and that is good advice for a league fixture too.
- The licence covers it. A backup PC is an add-on to the same licence, priced on the pricing page with everything else.
- It does not make the wall immortal. This protects against the machine that plays the wall. A dead cabinet, a cut cable or a processor that lost its input is on the other side of the DVI socket, which is what LED monitoring is for.
The setup, in one table.
- Machines
- Two Windows PCs, both cabled to the LED processors
- Network
- Both on the venue LAN, no internet needed for the failover itself
- Heartbeat
- One second, machine to machine
- Failover
- About five seconds of silence, then the backup is on the wall
- Switch back
- Manual, between halves or after the match, never automatically mid-fixture
- Screens
- Mirrored by position, so multi-screen grounds stay in step
- Licence
- A backup PC add-on on the same licence
- Power
- Separate circuits, and a UPS each if the fixture is a European one
What we have not solved.
There is a gap, and it is about five seconds long. If a sponsor rotation was due to change inside it, it changed late.
We could make the gap shorter by treating a single missed heartbeat as a failure, and we do not, because a machine that is briefly busy would then hand the wall away in the middle of a match for no reason. Five seconds of a held frame is the cheaper mistake.
And the switch back is deliberately yours. Software that fails back on its own would produce a second transition at a moment nobody chose, usually while somebody is still working out what happened to the first machine.
Ask us what happens in your ground.
Tell us how your two machines are cabled and powered. The weak point is usually in that answer, and it is not always the software.