All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
IPv6: Too big for routing tables?
Kaulkwappe
Member, LIR
I've been working with a datacenter for many years now. They have been always great, it really makes fun to work with them because they're very fast and the performance always has been stable over the years.
But there is one thing where they might not be so great: IPv6.
I have to say that I'm not so familiar with network or routing things at all because I am more a software developer. A few days ago I've asked them to announce a big IPv6 subnet and route a /48 into my VLAN so I can use the /64s from it on my dedicated root servers. They said something like that: No, they cannot do that because IPv6 has so many IPs and that would cause a high overhead. They are only able to route /64s and if I want to use larger subnets like /56s or /48s they only can provide me that over Next Hop Routing.
So my questions are:
a) Is that plausible? Do they possibly use too old hardware which is simply not able to handle IPv6 in an efficient way? But why are they able then to route IPv6 which already has an extreme amount of single IP addresses?
b) If they do Next Hop Routing for different /48 with different targets, can I handle this directly over my dedicated servers which would be the target, or do I necessarily need a router (e.g. a server with pfSense)? I want to avoid unnecessary overhead.
Thanks for your help!
Comments
Hahaha! Is it plausible? Of course not! EVERY router, no matter how poor, should be capable to have potentially the equivalent of todays complete IP4 address space TO THE SQUARE in its tables. Times 4 again, of course, because those IPv6 addresses are 4 times the length of IP4 addresses.
IPv6, the never ending source of joy ...
It's of course reasonable understandable. Imagine how much load you would make on the routing engine (control plane) and also on the FIB by installing one route for every single RIB entry you generate with a ipv6 neighbor.
You can easily fill up the TCAM space of a ISPs router, if you use all the addresses within a /64. JunOS for example has a configurable option to limit the usage of nd entries for specific interfaces / vlans.
Hi, unfortunately that's plausible yes. Routing for instance a /29 over a single vlan can cause a router to give issues in the distribution of the IPS as it's so many IPS. Really depends on the router model and what else is configured within the network.
What they could do is split the /48 up in /64s and route those to your vlan. Might help.
Might also be reminiscent of this issue with archaic dumb network equipment, or just poorly-educated people who are totally clueless when it comes to IPv6 stuff (alas! there are still many around, these days).