on 27-03-2010 18:11
on 10-07-2019 22:08
@ksim wrote:I am using Mikrotik, the configuration was done using the console, standard "6to4 Tunnel" followed the documentation on HE site, no shaping is configured on the router (did a clean setup), VM hub is in modem mode. CPU usage on my Mikrotik as about 0%, the same configuration worked fine with other providers.
my traceroute:traceroute -w 1 -n 216.66.88.98 1 192.168.1.2 0.895 ms 0.245 ms 0.240 ms traceroute to 216.66.88.98 (216.66.88.98), 64 hops max, 52 byte packets 2 * * * 3 62.255.32.121 11.692 ms 10.841 ms 10.045 ms 4 * * * 5 * * * 6 * * * 7 62.254.42.174 22.375 ms 15.433 ms 15.507 ms 8 * * * 9 84.116.132.5 28.421 ms 27.805 ms 30.051 ms 10 84.116.130.102 36.167 ms 28.854 ms 27.145 ms 11 216.66.87.125 36.936 ms 28.858 ms 28.143 ms 12 184.105.80.37 48.856 ms 27.011 ms 28.013 ms 13 216.66.88.98 28.061 ms 27.859 ms 26.987 ms
I am using a very similar setup with a MikroTik CCR1009, but with VM business. I migrated from residential to businesses, just the modem was swapped same cable etc.
I am also seeing this throttling. It has been happening for more than 4 months for me, when I was on residential and has not changed since.
I am going to try to find out which hop the throttling occurs on.
TonyJr
11-07-2019 09:26 - edited 11-07-2019 09:31
Here's the output of traceroute to my tunnel endpoint (216.66.80.26):
traceroute 216.66.80.26 traceroute to 216.66.80.26 (216.66.80.26), 30 hops max, 38 byte packets 1 10.167.160.1 (10.167.160.1) 12.523 ms 9.657 ms 9.918 ms 2 leic-core-2b-xe-830-0.network.virginmedia.net (62.252.13.254) 11.755 ms 12.652 ms 12.017 ms 3 * * * 4 * * * 5 * * * 6 * * * 7 * * * 8 de-fra04d-rc1-ae56-0.aorta.net (84.116.132.5) 29.838 ms 29.711 ms 29.982 ms 9 fr-par02b-rd1-ae102-0.aorta.net (84.116.130.102) 29.890 ms 29.758 ms 30.965 ms 10 10ge10-2.core1.fra1.he.net (216.66.87.125) 28.937 ms 29.760 ms 29.856 ms 11 100ge6-1.core1.lon2.he.net (184.105.80.37) 31.989 ms 32.701 ms 34.981 ms 12 tserv1.lon1.he.net (216.66.80.26) 30.874 ms 32.746 ms 35.857 ms
Also seems to be dancing through hops in France and Germany, if the names are to be believed 🤔
FWIW, if I traceroute to the same HE endpoint from work (on JANET) the traffic goes straight from ja.net to he.net via peering in London. I think VM's routing to HE may be a little screwy, or they simply don't have any peering in the UK with them?
on 11-07-2019 12:42
I thought JANET now used VM business, maybe not.
on 11-07-2019 13:10
JANET may purchase transit from VM but they also peer directly with many many other networks. They have a direct peering relationship with HE as you can see here: https://bgp.he.net/AS786
The route VM traffic to going to HE is very odd. It seems to be because they get their upstream from Core Backbone Gmbh (in Germany) as part of the Liberty Global group. They in turn appear to peer in Paris with Hurricane Electric.
Have prodded the HE looking glass server (web traceroute)(https://lg.he.net/) and it seems that traffic can go directly from their facilities in London to my VM IP. This looks like some specific traffic engineering or a configuration **bleep** up.
on 11-07-2019 14:15
11-07-2019 15:17 - edited 11-07-2019 15:31
Indeed, if I traceroute (IPv4) to my HE endpoint from work (JANET) it's a much more direct route than over IPv4 from home (Virgin Media). VM and HE are both present in LINX LON1, and JANET peers all over the place. Whacky routing by VM.
(When I say "more direct", it's only one more hop, but it's not going half way around Europe as part of it).
I wonder what difference it would make using HE's French or German tunnel endpoints? Time to experiment....
on 11-07-2019 15:53
Okay, if I'm reading this right, "de-fra04d-rc1-ae56-0" has 'de' signifying Germany and 'fra' for Frankfurt. Likewise, 'fr' is for France and 'par' is Paris on the VM/Liberty/aorta side of things.
traceroute from VM, Leics to Frankfurt tunnel endpoint:
traceroute to 216.66.80.30 (216.66.80.30), 30 hops max, 38 byte packets 1 10.167.160.1 (10.167.160.1) 12.181 ms 9.622 ms 8.064 ms 2 leic-core-2a-xe-823-0.network.virginmedia.net (62.252.13.213) 8.880 ms 9.737 ms 9.947 ms 3 * * * 4 * * * 5 86.85-254-62.static.virginmediabusiness.co.uk (62.254.85.86) 22.562 ms 19.373 ms 20.870 ms 6 * * * 7 de-fra04d-rc1-ae56-0.aorta.net (84.116.132.5) 32.373 ms 32.800 ms 35.979 ms 8 fr-par02b-rd1-ae102-0.aorta.net (84.116.130.102) 37.874 ms 30.552 ms 32.087 ms 9 10ge10-2.core1.fra1.he.net (216.66.87.125) 29.961 ms 31.679 ms 29.924 ms 10 tserv1.fra1.he.net (216.66.80.30) 29.954 ms 32.704 ms 38.045 ms
...so this loops *through* Paris.
traceroute from VM in Leics to Paris tunnel endpoint:
traceroute to 216.66.84.42 (216.66.84.42), 30 hops max, 38 byte packets 1 10.167.160.1 (10.167.160.1) 13.486 ms 9.700 ms 16.997 ms 2 leic-core-2a-xe-823-0.network.virginmedia.net (62.252.13.213) 8.867 ms 10.688 ms 9.963 ms 3 * * * 4 * * * 5 86.85-254-62.static.virginmediabusiness.co.uk (62.254.85.86) 23.478 ms 19.851 ms 22.928 ms 6 * * * 7 de-fra04d-rc1-ae56-0.aorta.net (84.116.132.5) 30.925 ms 30.713 ms 30.911 ms 8 fr-par02b-rd1-ae102-0.aorta.net (84.116.130.102) 36.001 ms 37.771 ms 29.882 ms 9 10ge10-2.core1.fra1.he.net (216.66.87.125) 29.958 ms 31.703 ms 29.980 ms 10 100ge1-1.core1.par2.he.net (72.52.92.13) 44.851 ms 34.756 ms 42.354 ms 11 tserv1.par1.he.net (216.66.84.42) 35.493 ms 42.699 ms 35.976 ms
...so you'd expect using the Paris endpoint would be closer, but no 🙄
I thought I'd traceroute all the HE tunnel endpoints in Europe from my VM connection. Everything transits that de-fra/fr-par pair, but Frankfurt is the nearest at 10 hops and Budapest and Warsaw are the furthest at 15 hops. I might mtr them overnight and get a better idea of latency along those routes...still doesn't address the traffic shaping issue, though.
11-07-2019 15:58 - edited 11-07-2019 16:02
one small thing, my upload speed is not shaped it went up with the plan update and bigger than download now, the cap might be caused by the short route back from HE to VM, rather the long way to, lower response times from the endpoint than 1 hop before is the evidence of this, maybe badly implemented DDOS protection somewhere.
2. ??? 3. brhm-core-2a-xe-022-0.network.virginmedia.net 0.0% 25 11.6 15.6 9.6 44.8 10.0 4. ??? 5. ??? 6. ??? 7. m686-mp2.cvx1-b.lis.dial.ntli.net 0.0% 24 13.8 16.5 13.5 30.5 3.9 8. ??? 9. de-fra04d-rc1-ae56-0.aorta.net 0.0% 24 29.8 29.4 26.1 37.8 3.4 10. fr-par02b-rd1-ae102-0.aorta.net 0.0% 24 28.6 28.7 26.3 35.4 2.3 11. 10ge10-2.core1.fra1.he.net 0.0% 24 26.5 29.1 26.1 36.3 3.6 12. 100ge6-1.core1.lon2.he.net 0.0% 24 27.3 30.6 25.4 61.4 7.6 13. tserv1.lon2.he.net 0.0% 24 25.3 26.2 24.3 34.2 2.3
on 11-07-2019 22:41
Interesting.
I've run a few traceroutes and VMs routing is odd. Using the HE looking glass server I get the following traceroute back to my IP:
core2.lon1.he.net> traceroute 94.173.xx.xx source 216.218.252.118 traceroute to 94.173.xx.xx (94.173.xx.xx), 30 hops max, 60 byte packets 1 100ge7-2.core1.lon2.he.net (184.104.192.133) 19.710 ms 19.433 ms 19.457 ms 2 lgi-upc-as6830.10gigabitethernet3-4.core1.lon2.he.net (216.66.80.118) 12.540 ms 12.556 ms 12.553 ms 3 * * * 4 uk-lon01b-ri1-ae25-0.aorta.net (84.116.136.102) 12.823 ms 13.014 ms 13.048 ms 5 * * * 6 sgyl-core-2a-ae11-0.network.virginmedia.net (62.254.42.149) 26.261 ms 25.564 ms 25.694 ms 7 sgyl-cmts-39-tenge414.network.virginmedia.net (62.252.46.122) 25.232 ms 25.288 ms 25.228 ms 8 94.173.xx.xx (94.173.xx.xx) 32.772 ms 31.199 ms 35.589 ms
If I go the other direction from my IP to the HE looking glass server:
traceroute 216.218.252.118 traceroute to 216.218.252.118 (216.218.252.118), 30 hops max, 38 byte packets 1 10.234.180.1 (10.234.180.1) 7.090 ms 10.206 ms 7.147 ms 2 sgyl-core-2a-xe-211-0.network.virginmedia.net (62.253.0.89) 7.848 ms 10.509 ms 8.800 ms 3 * * * 4 86.85-254-62.static.virginmediabusiness.co.uk (62.254.85.86) 20.546 ms 19.560 ms 19.900 ms 5 * * * 6 de-fra04d-rc1-ae56-0.aorta.net (84.116.132.5) 31.881 ms 44.774 ms 31.687 ms 7 fr-par02b-rd1-ae102-0.aorta.net (84.116.130.102) 34.851 ms 33.040 ms 31.759 ms 8 10ge10-2.core1.fra1.he.net (216.66.87.125) 31.671 ms 31.542 ms 31.989 ms 9 100ge6-1.core1.lon2.he.net (184.105.80.37) 34.437 ms 31.301 ms 33.263 ms
So going out goes via the Liberty Global European transit connection but coming in comes via a direct peer with Hurricane Electric. That's some interesting traffic engineering going on there. Perhaps they are trying to balance traffic levels or some other ISP black magic that is beyond my ken.
Perhaps it's time for those affected to take this up in a new thread on the "Speed" forum to get it escalated?
on 30-08-2019 09:03
so we're getting on for two months now with no further update...