There are two pieces here.
We do hypothesize (and design for) having competing EID allocators and
the ability for someone who got an allocation to move to a different EID
allocator covering the same space. That is necessary to keep the prices
for allocation and map maintenance competitive.
PITRs can be run by anyone who wants to (with suitable authentication /
authorization). So they can be run (as Dino hypothesizes) by the ISPs.
They can be run by the EID allocators. Or by third parties. In all
cases, they either have to be prepared for too much traffic or restrict
what they advertise into the BGP system. Which is why I hope that it
will make business sense for the allocators to run PITRs, since they are
the only ones in the right place to easily provide aggregation of EID
allocations.
Yours,
Joel
PS: Some of this is discussed in the lisp-deployment draft. Some of it
isn't.
On 12/4/13 3:03 PM, Ronald Bonica wrote:
Hi Joel,
In that model, assume that I obtain EID space from an EID allocator. At some
point in the future, the PITR service that that EID allocator provides becomes
inadequate. What is my can I do?
Are there competing EID allocators? If I go with a competing EID allocator, do
I have to renumber?
Ron
-----Original Message-----
From: Joel M. Halpern [mailto:[email protected]]
Sent: Wednesday, December 04, 2013 12:02 PM
To: Ronald Bonica; Dino Farinacci
Cc: LISP mailing list list
Subject: Re: [lisp] WGLC draft-ietf-lisp-eid-block-07
A partial answer that has been suggested is that the oeprator who
deploys the PITR is an EID allocator rather than a conventional /
existing ISP. Unfortunately, if LISp succeeds it is not clear that the
compensation levels can match the increasing demand for traffic in the
critical periods.
Yours,
Joel
On 12/4/13 11:51 AM, Ronald Bonica wrote:
Dino,
I am not understanding your response. Let me ask the question another
way.
Assume that an operator deploys a PITR. What policy can that operator
enforce to ensure that it is compensated for all (or even most) of the
traffic that it carries across that PITR?
Ron
-----Original Message-----
From: Dino Farinacci [mailto:[email protected]]
Sent: Wednesday, December 04, 2013 11:44 AM
To: Ronald Bonica
Cc: Luigi Iannone; Geoff Huston; Sander Steffann; LISP mailing list
list
Subject: Re: [lisp] WGLC draft-ietf-lisp-eid-block-07
Luigi,
Is this really what is going to happen?
If a PITR announces the entire /32 into the global Internet, it
puts
itself on the forwarding path for the entire /32, and incurs the
cost
associated with transporting traffic towards every site in that /32.
This is supportable only if the PITR operator is somehow compensated
for carrying all of that traffic.
But maybe only from a few sources. But if the /32 needs to be
divided
based on region, then maybe /40s could be advertised. But to the
point about "few sources", the more PITRs there are, the better the
load is shared.
And I envision PITRs will be deployed on on-path boxes anyways.
Those
boxes right now can route to the entire Internet, they are called PE
boxes, are they not Ron?
Isn't it more likely that the PITR operator will advertise only
slices of the /32, with each of those slices being assigned to
either
its customers (from whom it collects revenue) or the customers of
other operators with whom it has made financial arrangements?
No it won't be that way. EIDs are provider independent. If you do
what we suggest, we make no forward progress.
Dino
Ron
_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp
_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp
_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp
