Re: [Lsr] Request WG adoption of TTZ

[Christian Hopps <[e-mail]>]

> On Jul 7, 2020, at 12:00 PM, Huaimo Chen <[e-mail]> wrote:
> 
> Hi Chris,
> 
>     Thank you very much for your questions.
>     My answers/explanations are inline below with prefix [HC].
> 
> Best Regards,
> Huaimo
> From: Christian Hopps
> Sent: Tuesday, July 7, 2020 8:08 AM
> To: Huaimo Chen
> Cc: Christian Hopps; [e-mail]; [e-mail]
> Subject: Re: [Lsr] Request WG adoption of TTZ
> 
> Hi Huaimo,
> 
> Can you speak to the differences of this with Area Proxy? They are similar solutions, right?
> 
> [HC]: There are some differences even though they looks similar.
> At first, the target to be abstracted in Area Proxy is different from that in TTZ.
> Area Proxy abstracts an existing area to a single pseudo node.
> TTZ abstracts a zone to a single pseudo node. A zone is a piece or block of an area.
> An area is different from a zone in general.

[As wg member]

If the operator needed a subset of an area abstracted they could split the area and then use area proxy on the subset though, right? Restricting things to exactly the area boundary as area proxy does seems like an elegant KISS approach; it has the advantage of harnessing all the existing tools and institutional knowledge on how IS-IS areas work. This in turn simplifies management and debugability I think. While TTZ adds some extra flexibility (cuts out the area split step) I don't know that the added complexity makes sense since splitting an area is basically the same operation as configuring a zone would be.

> Secondly, the ways in which they are applied to an area for scalability are different.
> When an area becomes bigger and bigger, we may have scalability issues. Using TTZ, we can abstract one or a few zones in the area to one or few pseudo nodes smoothly without any interface down. Using Area Proxy, we need split the existing area into multiple areas, and then abstract one or a few areas to one or few pseudo nodes.
> These differences will produce different user experiences.
> For splitting an existing area into multiple areas, users may need put more efforts since it causes service interruptions in general. While splitting an area such as an OSPF area into multiple areas, the area numbers configured on some interfaces will be changed and each of these interfaces will be down with its old area number and then up with its new area number. These interface downs and ups will cause service interruptions in general.
> For defining zones in an area, users may need less efforts since abstracting a zone to a single pseudo node is smooth without any interface down.
> 
> Moreover, TTZ provides smooth transferring between a zone and its single pseudo node. That is that a zone can be smoothly transferred to a single pseudo node, and the pseudo node can be smoothly rolled back to the zone.

This strikes me as the important difference from area proxy. It certainly adds complexity to things, I wonder if it's worth it?

Thanks,
Chris.
[As WG member]

> 
> BTW, In the Area Proxy draft, Area Proxy abstracts an existing IS-IS area to a single pseudo node.
> In the TTZ draft, TTZ abstracts a zone in an OSPF area to a single pseudo node, and a zone in an IS-IS area to a single pseudo node.
> 
> 
> There's an existing experimental track OSPF RFC (RFC8099) already for TTZ so i found it confusing to have this document also talking about TTZ for OSPF; is it redefining it, updating it, just referring to it?
> 
> [HC]: Regarding to TTZ for OSPF, OSPF RFC (RFC8099) describes a solution for abstracting a zone to the zone edges full mess. This document proposes a new solution for abstracting a zone to a single pseudo node. The new solution re-uses some of RFC 8099, to which are referred. The new extensions to OSPF for the new solution are defined in the document.
> 
> Thanks,
> Chris.
> [chair hat]
> 
> 
> > On Jun 18, 2020, at 11:38 PM, Huaimo Chen <[e-mail]> wrote:
> >
> > Hi Chris and Acee, and everyone,
> >
> >
> >
> >     I would like to request working group adoption of "Topology-Transparent Zone"
> >
> > (TTZ for short) https://datatracker.ietf.org/doc/draft-chen-isis-ttz/ .
> >
> >
> >
> >     This draft comprises the following solutions for helping to improve scalability:
> >
> >         1) abstracting a zone to a single pseudo node in IS-IS,
> >
> >         2) abstracting a zone to a single pseudo node in OSPF,
> >
> >         3) abstracting a zone to zone edges' full mess in IS-IS, and
> >
> >         4) transferring smoothly between a zone and a single pseudo node.
> >
> >     A zone is a block of an area (IS-IS L2 or L1 area, OSPF backbone or
> >
> > non-backbone area).
> >
> >
> >
> >     When a network area becomes (too) big, we can reduce its size in the sense
> >
> > of its LSDB size through abstracting a zone to a single pseudo node or
> >
> > abstracting a few zones to a few pseudo nodes.
> >
> >
> >
> >     While a zone is being abstracted (or transferred) to a single pseudo node,
> >
> > the network is stable. There is no or minimum service interruption.
> >
> >
> >
> >     After abstracting a few zones to a few pseudo nodes, if we want to reconstruct
> >
> > them, we can transfer (or roll) any of the pseudo nodes back to its zone smoothly
> >
> > with no or minimum service interruption.
> >
> >
> >
> >     We had a prototype implementation of abstracting a zone to zone edges' full
> >
> > mess in OSPF. The procedures and related protocol extensions for transferring
> >
> > smoothly from a zone to zone edges' full mess are implemented and tested.
> >
> > A zone (block of an OSPF area) is smoothly transferred to its edges’ full mess
> >
> > without any routing disruptions. The routes on every router are stable while
> >
> > the zone is being transferred to its edges' mess. It is very easy to operate
> >
> > the transferring.
> >
> >
> >
> >     There are two other drafts for improving scalability: "Area Proxy for IS-IS"
> >
> > (Area Proxy for short) and "IS-IS Flood Reflection" (Flood Reflection for short).
> >
> >
> >
> >     "Area Proxy" https://tools.ietf.org/html/draft-li-lsr-isis-area-proxy-03
> >
> > abstracts an existing IS-IS L1 area to a single pseudo node.
> >
> >
> >
> >     "Flood Reflection" https://tools.ietf.org/html/draft-przygienda-lsr-flood-reflection-01
> >
> > abstracts an existing IS-IS L1 area to its edges' connections via one or more
> >
> > flood reflectors.
> >
> >
> >
> >     We believe that TTZ has some special advantages even though
> >
> > Area Proxy and Flood Reflection are very worthy. We would like
> >
> > to ask for working group adoption of TTZ.
> >
> >
> > Best Regards,
> >
> > Huaimo
> >
> > _______________________________________________
> > Lsr mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/lsr
>