[c-nsp] eliminating asymmetric routes

David Barak thegameiam at yahoo.com
Tue Aug 16 15:02:20 EDT 2005


In-line

--- Joe Maimon <jmaimon at ttec.com> wrote:

> 
> 
> David Barak wrote:
> > 
> > --- Joe Maimon <jmaimon at ttec.com> wrote:
> > what routing protocol are you using?  
> Whichever will do the job most naturally/simply on a
> cisco platform.

So EIGRP or OSPF are your friends :)

> > 
> > Are all of the below routers under your
> administrative
> > control?  Are the metrics symmetric in all cases? 
> 
> > 
> They will be. Metrics are symmetric and equal cost.


> > If the answers to the above two questions are
> "yes,"
> > then where is the asymmetry coming from?  
> If you count the hops, as a distance vector, A<->H
> have two equal cost 
> paths. A->C->E->F->H and H->F->D->C->A would be the
> asymmetricity.

Hmm - below you've got two diagrams, and the lower one
would imply that the shorter path would be A->C->F->H,
and the reverse path would be the shortest from H->A.



> > 
> > You mentioned that "warm standby routing" is not
> > preferred - do you mean that you want packets
> going
> > from C -> F to traverse both E and D, or would you
> be
> > willing to accept determinism in that case?  
> I want the shortest path between any two points, but
> the return path to 
> be guaranteed identical in reverse.

Ok good - I've seen some requirements for 50%
load-sharing but no asymmetric routing... needless to
say that was not pleasant...


> > 
> > Also, where are the NAT and FW boundaries which
> are
> > causing the constraint?
> > 
> > 
> C<->E
> C<->F
> D<->F
> D<->E

If those are NAT/FW boundaries, why are you routing
across them?  Why not have them be seperate routing
domains?

If you must, and prefer the first topology, however,
try this: set each link cost to "10" except for D-F,
which is 11.  Use OSPF, and you'll not have any
asymmetry.  If you want a decent open-source SPF
calculator, let me know off-list.

-David

> > 
> > 
> >>A   B
> >>|   |
> >>C---D
> >>|   |
> >>E---F
> >>| | |
> >>G | H
> >>   |
> >>   I
> >>
> >>Or
> >>
> >>
> >>A   B
> >>|   |
> >>C---D
> >>|\ /|
> >>| \ |
> >>|/ \|
> >>E---F
> >>| | |
> >>G | H
> >>   |
> >>   I
> >>
> >>E,F,I are all on a common subnet/vlan.
> >>
> >>Asymmetricity is a problem because of FW with
> >>natting OR without natting.
> >>
> >>This would occur on links
> >>
> >>C<->E
> >>C<->F
> >>D<->F
> >>D<->E
> >>
> >>The second design, while harder to scale
> eliminates
> >>inefficiencies.
> >>
> >>Warm standby routing (where only one link was used
> >>for ALL sites e.g. 
> >>C<->E is very much not preferred.
> >>
> >>So far on the table is-
> >>
> >>A)
> >>Extending connections (ethernet vlan/subnets) to
> >>ensure that all sites 
> >>have a one hop to all other sites across the
> nat/fw
> >>boundaries. 
> >>Difficult to scale.
> >>
> >>B)
> >>
> >>Tagging routes that cross
> >>
> >>
> >>C<->E
> >>C<->F
> >>D<->F
> >>D<->E
> >>
> >>And announcing them with higher cost on links
> >>
> >>C<->D
> >>E<->F(,I)
> >>
> >>Any ideas welcome.
> >>
> >>Joe
> >>
> > 
> > 
> > -David
> > 
> > David Barak
> > Need Geek Rock?  Try The Franchise: 
> > http://www.listentothefranchise.com
> > 
> > __________________________________________________
> > Do You Yahoo!?
> > Tired of spam?  Yahoo! Mail has the best spam
> protection around 
> > http://mail.yahoo.com 
> > _______________________________________________
> > cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/cisco-nsp
> > archive at
> http://puck.nether.net/pipermail/cisco-nsp/
> > 
> > 
> 


David Barak
Need Geek Rock?  Try The Franchise: 
http://www.listentothefranchise.com

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


More information about the cisco-nsp mailing list