Re: [Bier] Questions on draft-eckert-bier-te-arch-01

[IJsbrand Wijnands <[e-mail]>]

Hi Caitlin,

> Studying the architecture document it is apparent that the authors assume that there is some magic sauce
> associated with the routing underlay associated with each sub-domain. But there is almost nothing mandated
> about a sub-domain that is different from an SI slice of a sub-domain.
> 
> The SI slices of one sub-domain must use the same routing underlay. But nothing says that multiple sub-domains
> cannot use the same routing underlay.

The double ‘not' is a little confusing, but I think what you are saying is that multiple sub-domain’s are allowed to map to a single underlay, which is correct and the intention of the draft to allow that. You can also assigned a different underlay for each sub-domain. As Tony explained, SI is an automatic mechanism for when the BFR-id overspills, and we create a new SI for it. The sub-domain is something that the network operator can configure to implement specific services.

> 
> The BFR-IDs are unique to the sub-domain, and an SI is a slice of those IDs. But since BFR-IDs are assigned by a
> mysterious process outside the scope of the document this does not seem terribly relevant unless you need a
> BIER domain to have more than 2**16 ids.
> 
> Is there some reason why you could not replace the SI concept with simply have more sub-domains that mapped
> to the same number of routing underlays?
> 
> You end up with at most 2**24 BFR-IDs in either case, and the same options for sizing the Bitstring.
> 
> Were the drafters assuming that a sub-domain was "expensive"? There's nothing in the draft that requires every
> sub-domain to be a full cost sub-domain with a unique routing underlay.

A sub-domain is expensive in the sense that you need to allocate a new label(range) for it, advertise it through the network using the IGP and it will provision a new BIFT in all the BIER routers. 

Thx,

Ice.