dear lazynet:
sorry to ask an end customer question here, but we're at a bit of a
useless point, with actual clue being blocked by an account rep.
i have at&t gig fiber at home and it seems to be a pd /64. works fine.
at the office, we have at&t business 1g fiber going into an arris.
behind it we have a cisco asa with five lan segments. only ipv4
is enabled.
i want to enable v6 in the office, but can not find out what at&t is
providing. i would hope something such as a pd /56 with a static
allocation. after all, we pay through the nose for five static v4
prefixes.
anyone been to this movie and care to divulge the plot?
thanks
randy
Yes, one particular plotline which can explain why docsis systems do this is that standard residential customers are provisioned using giant broadcast domains directly on the cable, with DHCP config. Obviously it's more complicated because it's docsis, but lemme handwave and say that this is the gist of it. Because you're dealing with giant broadcast domains, you assign IP address blocks to individual CMTSs and your customers are assigned out of those ranges.
Assigning ipv6 in this context is really simple: it's part of the baseline DOCSIS3.0 standard and is supported incredibly well by all parts of the network.
Static addresses don't fit into this paradigm because you if you configure your static customers from a single broadcast domain, then they are glued to a particular CMTS and can't be moved from that CMTS unless you renumber them.
This doesn't work in practice because if you want to grow your network, you probably want to be able to move around chunks of your cable network from one CMTS to another in order to balance out your traffic. Or you might want to split a bunch of cable nodes from one CMTS to multiple, according as your traffic outgrew the capabilities of a single CMTS (a node in this context is a small chunk of a cable network).
One way of getting around this mess is to backhaul all your static customer interfaces using mpls l2vpn PWHE up to a L3 box which just handles static IP addresses. You configure the customer's static default gateway IP address on an interface on this head-end router, and the customer's cable modem will have a virtual connection directly to that interface. The thing is, this virtual interface termination system might or might not be tied into the entire ipv6 provisioning system. If it isn't, you're SoL. So even if dirt-cheap residential customers can get ipv6 very easily, it's different by virtue of the fact that you're using static IP addresses, because they're a headache for cable operators.
Nick
Yes, one particular plotline which can explain why docsis systems do
this is that standard residential customers are provisioned using
giant broadcast domains directly on the cable, with DHCP config.
Obviously it's more complicated because it's docsis, but lemme
handwave and say that this is the gist of it. Because you're dealing
with giant broadcast domains, you assign IP address blocks to
individual CMTSs and your customers are assigned out of those ranges.
Assigning ipv6 in this context is really simple: it's part of the
baseline DOCSIS3.0 standard and is supported incredibly well by all
parts of the network.
Static addresses don't fit into this paradigm because you if you
configure your static customers from a single broadcast domain, then
they are glued to a particular CMTS and can't be moved from that CMTS
unless you renumber them.
This doesn't work in practice because if you want to grow your
network, you probably want to be able to move around chunks of your
cable network from one CMTS to another in order to balance out your
traffic. Or you might want to split a bunch of cable nodes from one
CMTS to multiple, according as your traffic outgrew the capabilities
of a single CMTS (a node in this context is a small chunk of a cable
network).
One way of getting around this mess is to backhaul all your static
customer interfaces using mpls l2vpn PWHE up to a L3 box which just
handles static IP addresses. You configure the customer's static
default gateway IP address on an interface on this head-end router,
and the customer's cable modem will have a virtual connection directly
to that interface. The thing is, this virtual interface termination
system might or might not be tied into the entire ipv6 provisioning
system. If it isn't, you're SoL. So even if dirt-cheap residential
customers can get ipv6 very easily, it's different by virtue of the
fact that you're using static IP addresses, because they're a headache
for cable operators.
aha! makes sense.
i'll settle for dynamic. if i need static internally, i can always do
nat66 :)/2
i do not want to play how hard can we make ipv6 deployment; just want to
enable it on a five-segment office lan.
but i am beginning to see that there may be a reason i am having
problems getting past an account rep.
randy
I will speak more generally as I don't have insight into that provider.
Last mile providers are working on ipv6 everywhere because ipv4 is expensive and so is CGN and MAP-T.
IPv6 can reduce the need for ipv4 addresses and translation technology.
In all likelihood the device your office is connected to is not yet ipv6 enabled but there is a date somewhere on a calendar belonging to an engineering group or two.
The sales rep probably doesn't have a clue what that date is which is why he isn't giving you information.
He simply doesn't have it. Nor is an engineering group likely to give it to sales until it is a planned maintenance.
Engineers don't like promising things they can't deliver.
Sales likes promising the moon if they know there is cheese available at some point in the future.
Mack
The only thing that's worked for me is refusing to sign agreements or following through on cancellations unless it's delivered, not promised.
~Seth
Completely agree with all you've said here about how DOCSIS works. Several cable operators (at least) are working to make it so that DHCPv6 assigns business customers the same IPv6 prefix every time. Yes, it's DHCPv6 instead of static configuration, but I think it works.
Additional question, though:
Randy said "at&t business 1g fiber going into an Arris"
As fiber, it'll be PON. If it were a traditional cable company, I'd guess DPOE (DOCSIS Provisioning Over Ethernet).
Except it's AT&T, and I don't know if they're provisioning DOCSIS over their fiber; I would think it's GPON, using the same infrastructure as their U-Verse product (fiber to the curb, DSL to the home). That used to be PPPoE and not DHCP, but my information may be out of date.
Lee
AT&T fiber goes into a PON, and then into an Arris BGW210.
(Yes, I have business fiber here at satchell.net)