OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
From: Sina Mirtorabi (sinaCISCO.COM)
Date: Mon Jul 23 2001 - 14:25:02 CDT

  • Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

    Curtis Call wrote:

    >
    > > imagine the following topology
    > >
    > >
    > > R1------area 1 -------R2----- area 2 ------
    > > \ /
    > > \ area 3 /
    > > \ /
    > > R3
    > >
    > > so if I understand your idea, R2 will generate a summary for area 2 into
    > area 1
    > > and area 3 and will add an Area _Path ( 2 ) then R1 receiving this summary
    > will
    > > generate a summary to area 3 with the Area_Path ( 3,2 ) also R1 being part
    > of area
    > > 3 receive the summary from R2 with Area_Path ( 2 )
    > > so what is the rule for R1 to install the summary for itself ? if it's based
    > on
    > > the cost you could easily go into routing loop since R1 might use the path
    > through
    > > area 3 but R3 use the path through R1.
    > > by having a common area 0 and considering only summary from backbone we will
    > avoid
    > > this situation.
    > > I guess your idea works fine for not generating a summary back into an area
    > if the
    > > area number is present in the Area_Path but we need some how to tell the ABR
    > which
    > > path they have to choose themselves
    > >
    >
    > It would use the same logic as BGP does. I.e. the shortest Area Path would be
    > chosen. In this case R1 would have two possible paths to the destination, one
    > through Area 1 and one through Area 3, both with an Area Path of 2 (R1 would
    > ignore any summaries created by itself, just as is the current practice). The
    > tie breaker for same length Area Paths could be the lower area number, in
    > which case R1 would send packets through Area 1 in order to reach Area 2.
    > This would not cause a routing loop since all routers in Area 1 will use the
    > Summary created by R2 to reach Area 2 (it has a shorter Area Path).

    if the criterion is based on the shortest Area_Path in this case we could have
    sup optimal routing

    ABR------ area 1----------------------------ABR
        \ /
            \ /
               area 2----------ABR ------- area 3

    imagine the path through area 1 will go through much more number of router and the
    cost would be much higher than the path through area 2 and 3
    with this we basically some how fall back to the hop count metric.
    contrary to BGP in which the shortest path ( optimal ) is not the main criteria
    and path choice is based on policy in IGP this is very important to have optimal
    routing

    Sina