"serverMime-Version: 1.0

[Richard Laager]

On 4/3/19 3:24 PM, Gary E. Miller via devel wrote:
>> Just curious.  What sort of setup are you using where IPv6 is
>> significantly better than IPv4?
> Beats me.  All I can see is different paths using traceroute, mtr, etc.
> 
> Not unusual for ISPs and backbones to route IPv4 and IPv6 over different
> hardware.  They don't tell me why...

My day job is at a (small) ISP, so I am familiar with this. This (that
IPv6 is fewer hops and/or lower latency/jitter than IPv4) is an
extremely common observation.

The general consensus is that the big ISPs are doing traffic
engineering, but only applying it to IPv4 (since that's where the bulk
of the traffic is). Thus, the IPv6 traffic tends to follow a more
"natural" path, where the IPv4 is adjusted for capacity (but ultimately
financial) reasons.

If this is actually what is happening, a corollary is that as IPv6
traffic continues to grow, we may eventually see a point where traffic
engineering is applied to both, and the IPv6 paths match IPv4 (i.e. get
worse).

There are other factors, though, too, like Hurricane Electric's big push
with IPv6 that leads them to have a really high degree of peering on
IPv6. So that can lead to shorter paths with IPv6.

-- 
Richard

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <https://lists.ntpsec.org/pipermail/devel/attachments/20190403/8bc091ef/attachment.bin>