Please add option to output running statistics

[akorn]

oping(8) currently outputs one line per packet, like ping(1), and summary statistics on exit.

It would be useful to have an output mode where it prints statistics concerning the last n packets, or the last t seconds; something like this:

TARGET: last n: x packets transmitted, x received, xx.xx% packet loss, RTT[ms]: min = xx, median = xx, p(95) = xx, max = xx

The output should be line-oriented and parseable.

Use case: I have a script that monitors which of my uplinks is usable. Currently I invoke oping periodically, but it takes a while each time to get statistically significant results on e.g. packet loss. If I could instead keep oping running and read its output from a pipe, I would always have up to date statistics.

[akorn]

Actually, a sliding average over several different numbers of packets would also be useful (like the system load average). Say, if the average loss over the last 1000 packets is 3%, I might still think the link is good; but if the average loss over the last 25 packets it 30%, it's just going bad.

So I'd use the longer-term average to mark the link as "good" but the short-term average to mark it as "bad".

[octo]

I'd argue that there are two feature requests in this:

1. Add running statistics to the oping output. (noping already has that)
2. Provide an option to limit statistics to the last n packets (currently considers all packets since start).

Both seem reasonable and I would be happy to merge such pull requests.

Since load average came up: load average uses exponential decay, not a sliding window. While exponential decay is much easier to implement, I'd argue that a sliding window is more correct and much easier to grasp. For example, if you looked at the last 100 packets and the hosts goes from 0% packet loss to 100% packet loss, you'd expect the packet loss average to go from 0% to 100% in one percent point steps. With exponential decay, that is not the case.