GNOME Bugzilla – Bug 712385
netclock: added new round-trip limit property
Last modified: 2013-11-27 07:39:12 UTC
Created attachment 259934 [details] [review] netclientclock patch for round-trip limiting The netclientclock currently assumes that the network is of good quality, does not delay the packets too much, transports them in order etc. These assumptions generally hold for wired local networks, but not for wireless ones. In an area with many Wi-Fi access points and clients, time packets with round trips of 10 seconds have been observed. These extremely late packets mess up the regression calculations, causing the clock's stability to degrade significantly. This patch introduces a limit for the round trip period. If a packet takes longer to return than the limit specifies, it is ignored. By default, the limit is set to zero, which disables the limiting. (It is very difficult to determine a "good" nonzero default value.)
Comment on attachment 259934 [details] [review] netclientclock patch for round-trip limiting Needs to be updated to latest git master, and the property should have a Since: 1.4 gtk-doc marker. Otherwise looks good
Created attachment 262796 [details] [review] netclientclock patch for round-trip limiting v2
Review of attachment 262796 [details] [review]: Please rename the parameter to round-trip-limit, and fix the commit comment. It should have a summary line like netclock: Add round-trip-limit parameter Sometimes, packets might take a very long time to return. Such packets usually are way too late and destabilize the regression with their obsolete data. On Wi-Fi, round-trips of over 7 seconds have been observed. If the limit is set to a nonzero value, packets with a round-trip period larger than the limit are ignored. I think it would make sense to unify this RTT check and the RTT averaging immediately after, ie set rtt = GST_CLOCK_DIFF(clock1, clock2) once and use it, and make it so the rtt_avg also goes to the outlier_observed label, and to reset the timeout to a shorter value there as well, to retry the soon
Created attachment 262871 [details] [review] netclientclock patch for round-trip limiting v3
Comment on attachment 262871 [details] [review] netclientclock patch for round-trip limiting v3 Pushed, thanks