[email protected]
[Top] [All Lists]

Re: composite link - candidate for respin, maybe

Subject: Re: composite link - candidate for respin, maybe
From: Curtis Villamizar
Date: Mon, 29 Mar 2010 22:43:15 -0400
In message <C7D63887.910E%[email protected]>
Tony Li writes:
> > Then why bother with link bundling?  This level of control has been part of 
> > TE
> > since the beginning.  Link bundling was introduced to trade information 
> > hiding
> > for scalability.
> Because we can gain a great deal of scalability by advertising vectors of
> metrics rather than explicit metrics per link.  So, rather than:
>     link 1
>         bandwidth 10gbps
>     link 2
>         bandwidth 40gbps
>     link 3
>         bandwidth 100gbps
> We have:
>     composite links
>         bandwidth 10, 40, 100
> This saves an enormous amount of overhead in TLV encoding.  If we can
> further only vectorize a subset of the metrics for the composite link (e.g.,
> IGP metric and link coloring might be common), we get further scaling
> advantages.  And on top of that, little or no loss of useful information.

In this case it shouldn't matter which component link is chosen if the
LSP is under 10G.  If the delay on one link is shorter then it would
matter to some LSP but not others.

It save a lot if there are dozens of links of each type.  Then we

      composite link (singular)
          J @ bandwidth 10 delay X
          K @ bandwidth 40 delay Y
          L @ bandwidth 100 delay Z

> > See 
> > http://datatracker.ietf.org/doc/draft-ietf-mpls-explicit-resource-control-bund
> > le/
> Perfect, thanks.

Quite a bit less than perfect.  A small step toward the requirements.

> Tony

rtgwg mailing list
[email protected]

<Prev in Thread] Current Thread [Next in Thread>