[email protected]
[Top] [All Lists]

RE: composite link - candidate for respin, maybe

Subject: RE: composite link - candidate for respin, maybe
From: "Mcdysan, David E"
Date: Mon, 29 Mar 2010 14:20:52 -0400
Hi John,

Comments in line.



> -----Original Message-----
> From: [email protected] [mailto:[email protected]] 
> On Behalf Of John E Drake
> Sent: Monday, March 29, 2010 11:57 AM
> To: Tony Li; [email protected]
> Cc: [email protected]
> Subject: RE: composite link - candidate for respin, maybe 
> Comments inline
> > -----Original Message-----
> > From: [email protected] 
> [mailto:[email protected]] On Behalf 
> > Of Tony Li
> > Sent: Sunday, March 28, 2010 10:38 PM
> > To: [email protected]
> > Cc: [email protected]
> > Subject: Re: composite link - candidate for respin, maybe
> > 
> > 
> > Hi Curtis,
> > 
> > > What we don't want to do (IMHO) is completely remove 
> aggregation of 
> > > the bundle (aka CL) advertisement.  If there are a dozen 
> identical 
> > > links and four that are differnt there is no need to pick exactly 
> > > one, just one from either set, or don't care.
> > 
> > My concern is that down the road, as resources get used, 
> the links will no
> > longer be identical.   Each will have its own unique usage 
> levels.  Global
> > planning and optimization can use this information to determine 
> > exactly how many LSPs and at what sizes can fit down the link.
> [JD] 
> 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.

This is a key point. 

As mentioned in the current draft and as one of the first operator
stories that I told in the meeting is the current practice is to keep
the links separate since link bundling aggregation loses too much
information. If latency is important, then setting the TE-metric
proportional to the latency meets some of the requirements. However, if
one wants to set the TE-metric as a funciton of something other than
latency, then the status quo does not allow some customer or internal
use requirements in terms of latency to be supported. 

Some form of composition (or bundling) in routing must give some benefit
in terms of scalability/ stability as compared with separate links. 

> > 
> > 
> > > If we don't put a bound on the number of sets, in the limit each
> > > component could be advertised with the added semantic 
> that there is a
> > > sharing option.  In that case the LSP may need to specify 
> the set of
> > > "groups" (as I have it above in the absense of a term) that are
> > > acceptable.  This could make for a very big ERO.  It should be
> > > adequate to indicate in the RRO that the bundle is being 
> used with a
> > > field indicating which member was currently in use (if nailed in
> > > place).  This is beyond requirements, so if we are OK 
> with the stated
> > > requirement, let defer the details to later.
> > 
> > I'm fine with that, but we may also want to specify the 
> component in the
> > ERO.
> [JD] 
> See 

Being able to identify the component link on which an LSP is placed (or
spread across) is I believe a requirement which we need to capture,
which it appears this draft is proposing a specific solution.

> > 
> > > OK.  So what is your objection?  Are you suggesting that 
> we need to
> > > restrict the load balance to look down N labels from the 
> top and no
> > > further?
> > 
> > No, I'm suggesting that we need a bit in the RESV to say 
> that this LSP
> > should be demuxed by inner labels or kept atomic.
> [JD] 
> In current link bundling an LSP is either assigned to a 
> specific component link or split across all of the component 
> links.  In the latter case this effectively means all or a 
> subset of the component links.
> If an LSP is assigned to a specific component link, RFC 5150 
> (http://datatracker.ietf.org/doc/rfc5150/) (a one hop LSP on 
> the assigned component link) and MBB can be used to move the 
> LSP non-disruptively to another component link  

minimal disruption. Some disruption (e.g., packet reordering) could
occur if the links have different latencies. 

rtgwg mailing list
[email protected]

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