Thanks to all who responded! My confusion was that I expected the fast
switching cache to be cleared for a given route if a new route was
installed into the routing table. So, let me see if I have this
straight:
Routing table contains a route for a.b.c.d. The first packet is
process switched no matter what switching mode is enabled. In the case
of fast switching this populates the route cache and any subsequent
packets will be forwarded using the cached entry--and here's where I was
confused--even if a subsequent equal-cost route is installed into the
RIB.
Only after the cache is invalidated by other methods will the the
router repopulate the cache with both entries because the
process-switched packet now has two choices available in the RIB.
Is that essentially correct? I was under the mistaken impression that
the cache was more closely linked with the route installation process.
Here is a related question that I'll have to check into: if a route is
cached but then subsequently disappears from the routing table, does the
router continue to route using the cached entry or does it invalidate
the cache immediately at that point?
CCO, here I come.
Thanks all,
John
>>> "Howard C. Berkowitz" 6/12/01 7:14:06 AM >>>
>After reading a practice test question and answer I'm confused about
the
>operation of fast switching, specifically when a route has already
been
>cached when a new equal-cost route is learned via a different
interface.
>For example:
>
>Route A learns of 192.168.1.0/24 via e0 with a metric of 1000. Fast
>switching is enabled so this route is cached. Then the router learns
of
>192.168.1.0/24 with a metric of 1000 via e1. My thinking is that the
cache
>would be invalidated and recreated with two entries but the test
engine
>answer stated that routing would not change because the route was
cached and
>the cache would not be invalidated.
>
>Any thoughts? I'd test this myself but at the moment I only have two
>routers at home. Do any of you have any experience with this?
>
>Thanks,
>John
>
If I understand the question, the answer is correct. Where you are
getting confused is the difference in load balancing between process
switching and fast switching. Essentially, your thinking is correct
if the interface were in process switching mode, which does
per-packet load balancing. I'll make the minor nit that process
switching has no cache, but looks things up directly in the RIB
(i.e., main routing table). Formally, per-packet is deterministic
rather than statistical load balancing.
Fast switching uses per-destination load balancing. It will only have
one cached entry to any destination; the load balancing comes as a
result of having many cached entries to many destinations.
Per-destination load balancing is statistical, not deterministic, and
indeed can get unbalanced with a small number of destination. CEF
uses source-destination pair load balancing, which is still
statistical but increases the number of choices and reduces the
probability that the load will be unbalanced onto one interface in a
group.
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=8167&t=8110
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
