[j-nsp] VPLS: site-range

[Saku Ytti]

On (2014-01-08 00:21 +0000), Tom Storey wrote:

Hi Tom,

> From the reading around I have done, the site-range basically
> indicates how many sites maximum can/should exist for a given VPLS.

Yes.

> This tells the routers how many labels should be reserved for
> conveying data between each PE that has a site that is part of a given
> VPLS.

No. Site range can be 64k, and not single label is assigned. Labels are
assigned on-demand, in label-block-size, to cover all seen sites, not all
possible sites.

> If I say "site-range 10" then I shouldnt/cant provision more than 10
> sites for that VPLS, because insufficient labels will have been
> reserved for conveying data between any PE routers beyond the first
> 10, and thus those sites will be unreachable.

It won't allocate labels for higher sites, correct. 

> Suppose I configure all of my VPLS instances with "site-range 10"
> initially, if I need to add more sites later, I just need to go back
> and re-configure each VPLS instance to increase the number
> appropriately?

Yes.

> The default site-range value seems to be somewhere up in the 65,000's
> which from my reading is bad, because 65k labels will be reserved for
> that instance, which could be pretty wasteful if the VPLS only has 3-4
> sites.

It's sane, I don't think it should be even configurable. But maybe there is
some obscure CsC like situation where there is partial trust between two
parties and you want to ensure no one is stealing VPLS sites without paying,
only thing I can come up with.
I would keep it at maximum value.

-- 
  ++ytti