I have used this command in a production environment for dial on demand
with multiple ISDN channels, when using the load-threshold command. I
found that in my situation, the first channel kicked in quickly, but
because of the default five-minute interval, it took some time for the
reported load on that single link to reach the threshold so that another
link started up. As I often wanted half a dozen or more channels, it took
far too long for enough channels to kick in - in the meantime, the other
channels were being completely overloaded.
Dropping the load-interval to 30 (the minimum) helped quite a bit for this
situation. It also increased the likelihood of channels flapping, which in
my case was a reasonable trade-off.
I have also changed it to 30 for short-term monitoring of brief peaks of
traffic - peaks that don't show up as being too nasty in a five-minute
weighted average, but are quite long enough to disrupt application traffic.
However, I agree with Howard that it's not something to change arbitrarily.
JMcL
----- Forwarded by Jenny Mcleod/NSO/CSDA on 24/12/2001 09:27 am -----
"Stefan
Dozier" To:
[EMAIL PROTECTED]
Subject: RE: to set the interval in '
show
Sent by: interface'
[7:29981]
nobody@groupst
udy.com
24/12/2001
06:08
am
Please
respond
to
"Stefan
Dozier"
Here's the command:
R5-2511(config-if)#load-interval ?
Load interval delay in seconds
If you're applying this in a lab environment and are just curious
about how the numbers will look with different intervals, go for
it. If you're actually trying to accomplish something meaningful
by changing the intervals, I would definitely consider Howard's
questions and comments below!
Stefan
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of
Howard C. Berkowitz
Sent: Sunday, December 23, 2001 1:11 PM
To: [EMAIL PROTECTED]
Subject: Re: to set the interval in ' show interface' [7:29981]
>Hi ,
>I remember someone from here saying we can set the interval from the load
to a
>shorter time instead of the 5 minutes.
>
>#show interface ...
>
>Tks
I vaguely remember that there is a setting to do this, put in after
customer demands and against the recommendations of pretty much
everyone in routing engineering.
What problem are you trying to solve with a shorter interval? How
would the results of the statistics change anything you are doing?
Hypothetically, if you used the shortest possible interval, that of
one instruction time, the utilization would either be 0% (the CPU is
idle) or 100% (the CPU is doing something). The five-minute weighted
moving average represents experience for spotting trends beyond
instantaneous peaks.
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=29995&t=29981
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
