[email protected]
[Top] [All Lists]

Re: question on microloops

Subject: Re: question on microloops
From: Curtis Villamizar
Date: Sun, 31 Jul 2005 23:24:40 -0400
In message <[email protected]>
Alia Atlas writes:
> Hi Alok,
> On 7/31/05, Alok <[email protected]> wrote:
> > Hi,
> > 
> > With regard to:
> > http://www.ietf.org/internet-drafts/draft-ietf-rtgwg-microloop-analysis-00.txt
> > 
> > If the head end "signals" the path to the "tail end",
> > a. how can loops be formed?
> > b. Would the path setup given in the example (link A-C) not fail?
> > 
> > Well ofcourse I know the draft does not make an assumption that paths are
> > explicitly setup, but just curious....
> The question is where does the head-end get the traffic.  For RSVP-TE
> LSPs, the make-before-break mechanism ensures that micro-loops don't
> form along the LSP.  However, it is still possible for them to form
> between separate LSPs.

In a network running RSVP/TE if properly configured (not a partial TE
mesh) there is no possibility of forming loops unless traffic falls
back to IP.

For example, if there is a full mesh of RSVP/TE tunnels within a given
IGP area, no RSVP/TE loops can formed within the area.  It is
perfectly reasonable to do this with IGP areas of about 100 routers.
Right now it seems that about 400 routers per area would be pushing
the limit.

With multiple areas running RSVP/TE microloops are only possible to
the extent that they a loop crossing multiple areas is at least
theoretically possible.  A multiarea IGP microloops seems highly
unlikely.  If hierarchical RSVP/TE is used such that there is scalable
edge to edge TE then even these highly unlikely multiarea microloops
are not possible.  btw- I don't know anyone that is doing hierarchical
RSVP/TE to get edge to edge TE.


Rtgwg mailing list
[email protected]

<Prev in Thread] Current Thread [Next in Thread>
  • Re: question on microloops, Curtis Villamizar <=