Slow IPv6 Routes

We’re trying to increase IPv6 adoption by migrating things. However, slow IPv6 routes don’t help. We tested things using Vultr’s network as they had servers is 15 countries. These are an abbreviated sample of the results.

LONDON
HOST: vm1359261.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
6. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 10.0% 10 269.5 269.6 269.2 269.9 0.2


PARIS
HOST: vm1359265.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
9. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 10.0% 10 295.1 295.0 294.8 295.2 0.1


FRANKFURT
HOST: vm1359264.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
7. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 10.0% 10 293.6 293.8 293.6 294.0 0.1


LOS ANGELES
HOST: vm1359259.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
7. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 10.0% 10 202.3 202.3 202.0 202.6 0.2


NEW JERSEY
HOST: vm1359257.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
13. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 10.0% 10 275.1 275.0 274.6 275.3 0.2


MIAMI
HOST: vm1417800.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
11. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 20.0% 10 273.4 273.6 273.3 274.0 0.2


While the ping times from the US/UK were expectedly slow and matched closely to the IPv4 latencies, what surprised me a little was the ping times from the APAC region.
The IPv6 traffic from Australia seems to be routed via Japan and Hong Kong before landing in Malaysia.

SYDNEY
HOST: vm1359266.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
7. ae-9.r20.sydnau02.au.bb.gin.ntt.net 0.0% 10 0.6 0.6 0.5 0.7 0.1
8. ae-13.r24.osakjp02.jp.bb.gin.ntt.net 0.0% 10 115.6 117.5 115.6 125.6 3.5
9. ae-7.r25.tkokhk01.hk.bb.gin.ntt.net 0.0% 10 162.8 162.6 162.2 164.7 0.8
14. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 10.0% 10 283.3 282.4 281.9 283.3 0.5


It was also interesting that the traffic from Japan took a different route than the ones from Australia. This turned out to be the shortest latency via Hong Kong.

TOKYO
HOST: vm1359267.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
6. 100ge6-1.core1.hkg1.he.net 0.0% 10 160.7 162.3 157.4 167.7 3.7
9. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 10.0% 10 198.1 197.9 197.4 198.2 0.2


We had expected better performance from our neighbour Singapore, but things were terrible.

SINGAPORE
HOST: vm4672605.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
5. ae-10.r20.sngpsi05.sg.bb.gin.ntt.net 0.0% 10 1.1 7.8 1.0 66.9 20.8
6. ae-6.r24.tkokhk01.hk.bb.gin.ntt.net 0.0% 10 36.0 36.3 35.9 38.6 0.8
11. 2001:e68:5083:6d93:7555:c0a5:2945:cbd9 10.0% 10 246.9 246.5 246.0 246.9 0.3


Contrast this with the IPv4 route to and from the same place in Singapore. It just makes me wonder why we bother to use IPv6 in the first place.

SINGAPORE
HOST: vm4672605.vultr.com Loss% Snt Last Avg Best Wrst StDev
1. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0
4. 63-218-229-129.static.pccwglobal.net 0.0% 10 1.3 1.4 1.3 1.6 0.1
7. 60.52.64.180 0.0% 10 19.0 18.7 17.2 19.3 0.7


It is more than a 13x difference in latencies. So, it seems that we’ll be sticking to IPv4 networks for now.