On Sat, Apr 5, 2014 at 5:08 PM, Chuck Anderson <[email protected]> wrote: > And IPv6 over the HE tunnel: > > root@cerowrt:~# sh betterspeedtest.sh -H netperf6.richb-hanover.com > Testing against netperf6.richb-hanover.com while pinging gstatic.com (60 > seconds in each direction) > ............................................................................ > Download: 21.56 Mbps > Latency: (in msec, 77 pings, 0.00% packet loss) > Min: 14.477 > 10pct: 15.469 > Median: 17.646 > Avg: 18.906 > 90pct: 23.540 > Max: 36.302 > ............................................................................ > Upload: 5.85 Mbps > Latency: (in msec, 76 pings, 0.00% packet loss) > Min: 14.589 > 10pct: 15.579 > Median: 18.156 > Avg: 18.323 > 90pct: 21.192 > Max: 25.282
That's pretty lame compared to your ipv4 results, but the length of the path looks the same... puzzling... How much further (or less far) is rich's box (traceroute6 -n netperf6.richb-hanover.com) on ipv6 vs ipv4 (traceroute -n ) I have certainly seen bottlenecks, excessive delay, and packet loss on he's gateways. An "mtr" might be revealing during the test for spotting packet loss further on the path. > > > On Sat, Apr 05, 2014 at 08:02:37PM -0400, Chuck Anderson wrote: >> Here are some betterspeedtest.sh results for 3.10.36-1: >> >> First, without SQM enabled: >> >> root@cerowrt:~# sh betterspeedtest.sh >> Testing against netperf.richb-hanover.com while pinging gstatic.com (60 >> seconds in each direction) >> ............................................................ >> Download: 52.39 Mbps >> Latency: (in msec, 61 pings, 0.00% packet loss) >> Min: 15.281 >> 10pct: 18.302 >> Median: 28.502 >> Avg: 32.891 >> 90pct: 56.776 >> Max: 74.282 >> ............................................................. >> Upload: 11.07 Mbps >> Latency: (in msec, 61 pings, 0.00% packet loss) >> Min: 15.341 >> 10pct: 18.669 >> Median: 82.480 >> Avg: 126.662 >> 90pct: 248.102 >> Max: 278.644 >> >> And now, with SQM set to 80% up/down numbers from above: >> >> root@cerowrt:~# sh betterspeedtest.sh >> Testing against netperf.richb-hanover.com while pinging gstatic.com (60 >> seconds in each direction) >> ............................................................ >> Download: 32.84 Mbps >> Latency: (in msec, 61 pings, 0.00% packet loss) >> Min: 15.623 >> 10pct: 16.077 >> Median: 17.634 >> Avg: 17.982 >> 90pct: 19.653 >> Max: 23.272 >> ............................................................. >> Upload: 8.25 Mbps >> Latency: (in msec, 61 pings, 0.00% packet loss) >> Min: 16.001 >> 10pct: 17.623 >> Median: 19.796 >> Avg: 19.820 >> 90pct: 21.716 >> Max: 23.228 >> root@cerowrt:~# >> >> >> On Sat, Apr 05, 2014 at 01:18:51PM -0700, Dave Taht wrote: >> > + openwrt merge >> > ++ fix for dhcpv6 renew problem >> > + actually tested for an hour so far on 5.4ghz, with a us countrycode >> > and wpa+psk enabled... >> > >> > Get it at: >> > >> > http://snapon.lab.bufferbloat.net/~cero2/cerowrt/wndr/3.10.36-1/ >> > >> > but: there isn't much other reason to upgrade to this... >> > >> > - no progress on the wifi bug - but I am beating up wifi with a variety of >> > devices and scripts today hoping to make it fail, and bringing up a >> > bunch more tomorrow. >> > >> > - toke's script relies on stratum '16' changing, and it doesn't with >> > openwrt's >> > ntp, it seems.... > _______________________________________________ > Cerowrt-devel mailing list > [email protected] > https://lists.bufferbloat.net/listinfo/cerowrt-devel -- Dave Täht Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html _______________________________________________ Cerowrt-devel mailing list [email protected] https://lists.bufferbloat.net/listinfo/cerowrt-devel
