Pardon my ignorance - what do you see missing in MPLS in regards to support for IP6?-------- Original message --------
MPLS requires an IPv4 core. You can't run an IPv6-only infrastructure because neither CSCO or JNPR have implemented LDP to distribute labels for IPV6 prefixes.
-mel via cell
Pardon my ignorance - what do you see missing in MPLS in regards to support for IP6?
You can still carry the v6 NLRIs in MP-BGP though right?
Joshua Moore
Network Engineer
ATC Broadband
912.632.3161 - O | 912.218.3720 - M
Yes. But the MPLS nodes must all connect via IPv4.
-mel via cell
You can still carry the v6 NLRIs in MP-BGP though right?
Joshua Moore
Network Engineer
ATC Broadband
912.632.3161 - O | 912.218.3720 - M
Not true - Cisco have it in IOS XR since 5.3.0.
Juniper expect to start shipping it later in 15.
Mark.
That's good to hear!
-mel beckman
We added LDP IPv6 support in SR OS 13.0.R1 for Alcatel-Lucent 7x50 platforms
earlier this year.
Regards,
Greg
Greg,
After investigating what a previous poster said about Cisco and Juniper, I'm getting the feeling that not all major impediments to running MPLS over IPv6-only networks have been addressed.
Your comment mentions LDP IPv6 support. Do you now handle all the major gaps identified the the IETF MPLS IPv6 Gap Analysis (RFC7439) from this last January?
https://tools.ietf.org/html/rfc7439#section-3
It seems like their are still gaps in the MPLS spec itself before IPv6 has parity with IPv4 in MPLS.
-mel beckman
The LDPv6 support is just the control plane portion to get labels
assigned to IPv6 addresses. This should get you basic forwarding of
encapsulation and forwarding of IPv6 traffic in MPLS. The immediate
use-case would be removal of IPv6 BGP routing in the core, if that is
your thing.
Otherwise, yes, there are still a bunch of MPLS gaps that need to be
fixed for those additional services to run natively over an IPv6-only
network. Baby steps...
Mark.
I think the “THING” that people are starting to worry about is how to deploy a network when you can’t get IPv4 space for it at a reasonable price.
Owen
Owen,
Paying for IPv4 space definitely raises the capital requirements for any new provider startup. It's not so bad right now, when deals are plentiful in the $10k to $20k range for /24s. But when a /24 hits $100K, bootstrapping a new ISP will be impossible.
-mel beckman
Tell a start-up ISP it'll be $10k - $25k for PI IPs and they'll laugh in your face.
I suppose the issue will become "more real" when you can't get any IPv4
space period.
Mark.
That will never happen. If you offer me $1000 per IPv4, then I will happily
terminate some user contracts and sell their IP space to you...
In fact it will never become even that expensive. With a marked price of
$10 I am buying IP space for customers as needed and I will include free
space in the contracts. If the price went to $100 I would tell all users
that they need to pay monthly rent for their IP or alternative, the user
would have to accept carrier NAT in some form. And then I would proceed to
buy a new house for the money I make by selling address space.
There is a ton of address space that is inefficient used. We will be able
to buy excess from companies that "create" space by optimizing their
existing space. There is a reason we have not seen any rise in the price
even after multiple years with depletion in large parts of the world.
Regards,
Baldur
In this particular case, I'm not concerned about the next ten years.
Predicting what happens between now and then could have a fair degree of
accuracy.
I'm more concerned about what happens beyond that. I'm not sure I can
accurately (even with large error margins) predict what happens then.
All that said, I'm not trying to paint myself into that kind of corner.
It is 2015, after all... Just don't tell my competitors...
Mark.
Yes, the reason is that we'd never had ARIN turn down a request due to space exhaustion before. In 12 months we'll see the prices will go up significantly. Don't underestimate the demand, which is easily measured via ARIN space allocation reports. That demand rate has very little flexibility, and the businesses asking for /21 and above are willing to pay for the space. It's not the "two guys and a router" startups asking for a mere /23 or /24. These are generally pre-existing businesses or well-funded startups.
-mel beckman
I suppose the issue will become "more real" when you can't get any IPv4
space period.Mark.
That will never happen. If you offer me $1000 per IPv4, then I will happily
terminate some user contracts and sell their IP space to you…
Eventually, you run out of user contracts to terminate.
In fact it will never become even that expensive. With a marked price of
$10 I am buying IP space for customers as needed and I will include free
space in the contracts. If the price went to $100 I would tell all users
that they need to pay monthly rent for their IP or alternative, the user
would have to accept carrier NAT in some form. And then I would proceed to
buy a new house for the money I make by selling address space.
Sure, but aren’t your customers going to start demanding IPv6 instead of that at some point?
Aren’t they going to start insisting on a service that doesn’t charge per address?
There is a ton of address space that is inefficient used. We will be able
to buy excess from companies that "create" space by optimizing their
existing space. There is a reason we have not seen any rise in the price
even after multiple years with depletion in large parts of the world.
Yes… It’s called “soft landing”… ARIN will be the first region to deplete without significant
austerity policies for newcomers to get address space.
Owen
> That will never happen. If you offer me $1000 per IPv4, then I will
happily
> terminate some user contracts and sell their IP space to you…
Eventually, you run out of user contracts to terminate.
At $1000 per contract I do not care. I will retire and be happy.
Seriously ISPs are often valued by number of active contracts. A typical
value might be $100 to $200. The value of an IPv4 address can not go any
higher than this because at that point you can buy another company just to
get the addresses.
> In fact it will never become even that expensive. With a marked price of
> $10 I am buying IP space for customers as needed and I will include free
> space in the contracts. If the price went to $100 I would tell all users
> that they need to pay monthly rent for their IP or alternative, the user
> would have to accept carrier NAT in some form. And then I would proceed
to
> buy a new house for the money I make by selling address space.
Sure, but aren’t your customers going to start demanding IPv6 instead of
that at some point?
My customers have been demanding IPv6 for some time already.
Aren’t they going to start insisting on a service that doesn’t charge per
address?
They will have their free IPv6 and will care less about a non shared IPv4.
This will cause the valuation of IPv4 to crash at some point because the
demand will disappear.
> There is a ton of address space that is inefficient used. We will be
able
> to buy excess from companies that "create" space by optimizing their
> existing space. There is a reason we have not seen any rise in the price
> even after multiple years with depletion in large parts of the world.Yes… It’s called “soft landing”… ARIN will be the first region to deplete
without significant
austerity policies for newcomers to get address space.
RIPE gives you 1k addresses. This is not enough even for two guys and a
router. But yes it is a nice token.
The big companies have now gone without new space for a while. Just a few
months ago I bought 2k addresses at $6 per address. I do not observe any
rising tendency in IPv4 pricing.
Regards
Baldur
one word. RFC 1918. Here is an perpetual well of IPv4, packed down, overflowing.
manning
bmanning@karoshi.com
PO Box 12317
Marina del Rey, CA 90295
310.322.8102