[email protected]
[Top] [All Lists]

Re: intra-domain diameter

Subject: Re: intra-domain diameter
From: Pierre Francois
Date: Thu, 04 Oct 2007 15:50:22 +0200
[email protected] wrote:
Between the simple TTL-mechanism and IPFRR I can imagine many different mechanisms to deal and cope with loops which require different amounts of complexity. But let me address a different but related aspect. How big can an OSPF network be? Could it be that its size would require a larger TTL start value than 255 ? Given this issue, would it make sense to "define" a diameter being the sum of the distance between the current (root) node and its most distant node X1 plus the distance to its second most distant node X2 which however must be beyond some root-adjacent link which is not the best for getting to X1 ? If this kind of diameter is smaller than 255 then there is no problem with the TTL size.
Otherwise, no safe statement can be made, further study is required.
Has anyone already made related investigations ?
What we investigated w.r.t. the TTL is that :
Upon IP FRR activation, LFA or Notvia, the end-to-end path are not much longer in terms of number hops than
1. The initial path hop length within the topology
2. The post-convergence path hop length within the topology.

This means that if you drop a packet due to TTL expiration when it goes through an IP-FRR protection : 1. There is a significant chance that the packet would have been dropped before the failure anyway. 2. There is a significant chance that a packet sent from the same source to the same destination with the same TTL will be dropped when
the network has converged.

This has been verified on topologies ranging from small national ISPs to damn huge Tier-1 ISP topologies.
The conclusion is that
Size doesn't matter, diameter does, BUT this diameter doesn't change much upon a link / node failure.


Heiner ------------------------------------------------------------------------
rtgwg mailing list
[email protected]

rtgwg mailing list
[email protected]

<Prev in Thread] Current Thread [Next in Thread>
  • Re: intra-domain diameter, Pierre Francois <=