Six Cameras, One Disk

Home video surveillance is not a computer vision problem. It is an arithmetic problem about bytes, and the arithmetic does not negotiate. Six cameras, running always, writing to one disk.

The first number that matters is what each camera actually emits, measured at the recorder rather than read off the camera's own settings page. A camera configured for four megabits will happily send six when the wind moves a hedge, because the encoder is chasing a quality target and the scene got harder.

Sum it, multiply by eighty-six thousand four hundred seconds, and the day costs about two hundred and forty gigabytes. Divide the disk by that and you have the only retention number worth quoting — the age of the oldest hour you can still produce.

Motion detection is the usual answer to that number and it is a bad one at home. Recording only on motion means the archive has holes exactly where the argument later is about whether something happened. Continuous recording with a shorter window is the honest trade: less history, but no minute of it missing.

Which makes uptime the metric, and uptime is not a number — it is a shape. A camera that dies quietly is worse than one that dies loudly, because the RTSP session closes, the recorder shrugs, and nothing complains until somebody goes looking for a specific evening.

So the dashboard is not a log. A log tells you what happened when you already know where to look. A grid of dots tells you, at a glance, that yesterday has a hole in it — and the useful question was never how many cameras are online right now, but whether every minute of the retention window can actually be produced on request.