The MPLS WG Archive

[David Charlap <david.charlap@marconi.com>]

Gaitonde Anandprasanna wrote:
> 
> If u can remeber there was some discussion going on
> How to fill ERO object in case of RSVP-TE , but somehow i was unable
> to ask my questions at that time.
> 
> I have the follwoing queries:
> 
> 3 options had been considered:
> 1) Put the router ID of the routers. This does not work with routers
>  having multiuple links and unnumbered interfaces.
> 
> First i would like to know how can we put the router ID in the ERO
> subobject. If we see the format of the ERO subobject we can put IPV4
> prefix, IPV6 prefix or AS Number inside that object .
> 
> so how can we put router id. Is it in IP address format???

Most routers choose one of their interface IP addresses as a router ID. 
Even if some other address is manually configured, I think routing
protocols advertise these addresses.

With respect to multiple links and unnumbered interfaces, you can still
specify an ERO, but not with full precision.  You can still specify a
list of routers that the LSP must go through, although you won't be able
to specify specific links.  Since router IDs (as defined by OSPF and
ISIS) are 4-byte values just like IP addresses, the existing IPv4 prefix
subobject should work.

> 2) Put in the egress interface of the routers on the path.
> That is  remote IP addrress of each link .
> 
> But in case of unnumbered links how can we get the IP address of the
> remote end.

Via a local routing table.  If the required information doesn't exist
(meaning it wasn't learned via routing protocols and was not
configured), then there is no way to determine the next-hop interface
and a PathErr will be generated.

> Refer RSVP-TE draft Section 4.3.4.1 :
> 
> SElection of Next Hop:
> 
> 5) Interior of the abstract nnode case:- Otherwise, the node slects 
> the next hop within the abstract node of the subobject ( which the
> node belongs to) that is along the path to  the abstract node of the
> second subobject ( which is the next abstract node).
> 
> How can we do this ?? Please let me know. (Perhaps i am asking an
> implentation question here...)

If an ERO subobject doesn't uniquely specify a single next-hop (could be
a loose subobject, an AS number, a prefix shorter than /32, or other
reasons) then the switch must pick a next-hop using whatever means it
has.  I think most implementations will use the switch's local routing
table, possibly taking QoS requirements into account.

If the switch doesn't have a local routing table, then it will obviously
be unable to process a a subobject where the next-hop is ambiguous and
will generate a PathErr.

-- David