[email protected]
[Top] [All Lists]

Re: hop count vs real metrics LFA analysis

Subject: Re: hop count vs real metrics LFA analysis
From: Michael Menth
Date: Wed, 12 Dec 2007 16:02:15 +0100
Hi Pierre,

thanks for this note. Just to explain the difference of the analyses to the list: * Pierre's numbers tell how many links can be fully protected by LFAs. A link is fully protected if ALL destinations carried over that link can be protected by some LFA, no matter which type, and the protecting LFAs do not need to be the same for the protected destinations (?). * TR432 ( http://www3.informatik.uni-wuerzburg.de/TR/tr432.pdf ) shows how many destinations can be protected at a node by which kind of LFA when a certain protection level is required. Increased resilience requirements reduce the set of applicable LFA types.
In any case, the existence of LFAs is very sensitive to the topology and
to the link metrics. Certainly, an analysis of the existence of
different LFA types in operational networks with real link metrics would
be interesting. I hope that our national grant will be extended such
that we can have a deeper look at that issue. In this case, we would
appreciate interest and willingness of network operators to cooperate
with us on this area.
Pierre, I hope we'll soon find some time to work together on some ideas
that we sketched out in Vancouver. It was great talking to you.


Pierre Francois wrote:


To illustrate the discussion in rtgarea about the importance of
using the configured link metrics vs using hop count when looking at LFA coverage, here's a comparison of LFA's coverage, considering the percentage of links that are fully covered by LFAs in various topologies.
It shows the coverage with configured metrics vs. coverage with hop
ISPs A,  D, and E are national ISPs. ISPs B and C are Tier-1 ISPs
Note that the results for ISPE have *not* been swapped.

ISPA : 46% vs 12%
ISPB : 50% vs 40%
ISPC :  66% vs 46%
ISPD : 31% vs 21%
ISPE : 7% vs  46%



rtgwg mailing list
[email protected]
Dr. Michael Menth, Assistant Professor
University of Wuerzburg, Institute of Computer Science
Am Hubland, D-97074 Wuerzburg, Germany, room B206
phone: (+49)-931/888-6644, fax: (+49)-931/888-6632
mailto:[email protected]

rtgwg mailing list
[email protected]

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