>At 09:54 AM 11/16/2001 -0500, [EMAIL PROTECTED] wrote:
>>I have implemented OSPF Load Balance by making the cost of the interface
>>equal. For some reason traffic is incomming traffic is prefferd on one of
the
>>links. I have bounced the adjacencies and OSPF process. before I twig the
>>cost to balance out the traffic has anyone had this issue?
>
>
>I don't think you accomplished what you wanted to do.  Setting all 
>the costs to be the same causes OSPF to act like RIP.  That is, the 
>metric now becomes a hop count.  So unless you have the same number 
>of hops, it won't work.

Yes and no. If, for example, it is a campus network where everything 
is 100 Mbps, using what is effectively hop count still gives accurate 
metrics.  In general, I find topology more important than exact 
metric values in finding optimal paths. Distribution and core speeds 
tend to be standardized.  Where the link speeds tend to vary 
significantly is at the access tier, where you might not have 
multiple paths in any case, or a very trivial selection such as 
dedicated versus dial.

>
>Even then, route-cache will cause per destination load-balancing.

True for route-cache only. No route-cache will give per-packet, and 
various flavors of CEF will give source-destination.

>
>One other thing to consider, you cannot control incoming packets 
>(unless you use bgp)..

But if all the routers in the domain control their outgoing packets, 
you get very close to accomplishing that. I agree that external 
information control can only be affected by BGP, and then it's 
advisory rather than mandatory.

There's a fundamental question of how far one wants to go in traffic 
control, and why.  If it is seriously necessary, IP load balancing 
just doesn't have the level of control needed. You want to go to a 
traffic engineering model with traffic-aware IP routing protocols 
and/or MPLS.

>
>hsb




Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=26578&t=26578
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]

Reply via email to