Calculation mechanism behind the Tunnel Latency Stats?

[NewtonNet]

Hi,

A question borne out of idle curiosity if I may... How are the figures calculated for the Tunnel Latency Stats page?

I understand that the latency is periodically measured from the relevent tunnel server - presumably by sending pings? If so, how many and when? Are multiple pings sent in succession to account for any pre-connectivity lookups (Layer 2 addressing) etc that would otherwise skew the results?

Again, just curiosity that's all..  :)

Mathew

[broquea]

Once a week, the client side of the v6 tunnel is pinged 5 times from the tunnel-server, and the average is the result stored. They are scheduled for early AM, Pacific Time.

[NewtonNet]

Ah okay, thanks!

Mathew

[snarked]

I would suggest that the pings not be exactly weekly, but every 6 or 8 days.  The reason for the variance is to avoid hitting the same weekly time and thus being skewed by any weekly cron jobs that [client] tunnel endpoint machines are running.  By being an even number, test time also slips around when we have odd months (months with an odd number of days).