>Billy Monroe wrote,
>
>An interviewer asked me what happens if you have two paths to a Router, both
>using OC-3, and routing protocols is OSPF.
>
>I said that OSPF is a link state protocol and its metric is "Cost", which is
>10^8/Bandwidth. Load balance is enabled by default on OSPF, so packets will
>be load balanced "per destination" (it should be configured if 'per packet'
>is required).
>
>The guy then told me that I should take "Congestion" into consideration.
You were right. He was wrong.
>
>How can this be ? I know that EIGRP or IGRP may be configured to use Load as
>a metric, thus congestion would be take into account. I couldn't find any
>documentation showing that OSPF takes congestion as a metric.
He is REALLY wrong in trying to consider congestion as part of a
routing algorithm. While (E)IGRP can consider it, EIGRP does so only
for IGRP compatibility. IGRP's use of load is one of those things
that sounded good at the time of development, but has not been useful
in practice.
There are emerging protocols that consider load/utilization, but in a
manner quite different from the way IGRP does. IGRP only considered
utilization on directly connected links, which can lead both to route
oscillation and very bad end-to-end routing.
One newer approach (e.g., OSPF-TE, ISIS-TE) is to do explicit
bandwidth reservation before the routing protocol runs. When the
routing computation runs, costs are derived from available bandwidth,
not link bandwidth. Bandwidth reservation is either static or over a
substantial time period (e.g., RSVP).
Another approach (OSPF-OMP, ISIS-OMP) does consider utilization, but
averaged over a significant period, and as part of a link state
algorithm that does consider end-to-end.
>
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
