[j-nsp] OSPF Metric in Juniper.

[Wayne E. Bouchard]

On Wed, Jan 21, 2004 at 05:24:45AM -0500, brad dreisbach wrote:
> On Wed, Jan 21, 2004 at 11:24:59AM +0100, Lars Erik Gullerud wrote:
> > On Wed, 2004-01-21 at 11:04, Milind Deshpande wrote:
> > 
> > [snip]
> > > When we issue ping from Router A to Router D, ideally as per OSPF metric
> > > which is cost, it should follow the "Router B & Router D" path, but it is
> > > following "Router C & Router D"
> > > What could be the reason of taking path which are having both STM1 links,
> > > instead of Gigabit & STM1 path?
> > 
> > Unless you have increased the OSPF reference bandwidth in your network
> > or manually assigned metrics to interfaces, anything >= 100Mbit has a
> > metric of 1 (100Mbit is the default reference-bandwidth for OSPF cost
> > calculation). Which means A-C-D and A-B-D are equal-cost paths (same
> > number of hops, all hops are metric 1). Then  router A will load-balance
> > between the two, a "show route <ip of D>" should show you both paths.
> 
> yea you'd think traffic would be load balanced...according to this anyways:
> http://www.juniper.net/techpubs/software/junos/junos61/swconfig61-routing/html/ospf-config17.html#1014485
> 
> i wonder what effect having traffic-engineering shortcuts enabled
> will have though.hmmmm...

Except that according to the documentaiton I've seen (and noted
behaviors from experience), even per-packet load balancing is per-flow
(after a fashion at least). So seeing it take only one path does not
surprise me.

---
Wayne Bouchard
web at typo.org
Network Dude
http://www.typo.org/~web/