Every subscription here runs on two servers, not one. Both are asked at the same time and whichever answers first is the one the receiver uses. That is the whole mechanism, and the rest of this is about what it means for you and for the customer sitting in front of the television.

Both servers are asked, every time

"Load balancing" usually means something else. Normally it means splitting the work up: each request goes to one machine out of several, and something in the middle decides which machine gets it.

That is not what happens here. Your customer's receiver has both servers in it. It asks both of them every single time, and it uses whichever answer comes back first. Nothing is shared out, and nothing in the middle is choosing.

The difference shows when one machine is having a bad moment. If the work were split up, anything sent to the slow machine would sit there waiting for it. Here there is nothing to wait for, because the other answer is already on its way.

How a request from a receiver reaches a card A receiver sends a request for a key. It goes to both servers of its pair at once. Whichever answers first is the one the receiver uses, so a server that is slow or down costs a moment rather than a picture. Your customer CCcam or OSCam asks asks First server answered in 310 ms Second server answered in 640 ms Local card the real subscription Local card the same provider first answer wins Both servers of the pair are asked at once. A slow one costs a moment, not a picture.

What a customer notices

Nothing, which is the point.

A channel change asks for a key and waits for it. If one of the pair is slow to answer, the other one has usually already answered, so the picture arrives on the same schedule it always does. If one of the pair stops answering entirely, the receiver carries on with the other one without being told to and without anybody touching the box.

That is why the file we generate has both servers in it, and why handing a customer only one line would be worse than useless: one line is a single point of failure pointed straight at somebody's living room on a Saturday night.

What we measure

What we publish is the thing a customer actually feels: how long a channel takes to open, and how often changing channel worked, taken from real viewing rather than from a test rig. Both are on the front page, and both are read from the panel when this site is built rather than typed in by hand.

Trying it

A 24 or 48 hour line costs nothing. Put one on your own receiver, on your own dish, at the hour you actually watch, and change channel a few times. That tells you more than any figure on any supplier's website, including this one.

cccam server