From: Wayne Gustavus (wgustavus@xxxxxxxxxxxxxx)
Date: Tue May 08 2001 - 12:49:45 GMT-3
I believe the proper way to interpret the syntax of the command is more
basic than what you described. The percentages that are used in the backup
load command are definitely in relation to the available bandwidth of the
primary interface. The description of the disable-load threshold is trying
to convey the idea that the *total* traffic must be less than the value
derived from the disable-load threshold before *all* traffic will be moved
back to the primary link and the backup link is shut down.
For example, suppose you have a primary link with a bandwidth of 100Kbps
defined and the command "backup load 20 10". (I know the numbers may be
unrealistic but it makes the math easy and is easy to test)
If there is 50Kbps of traffic on the primary interface then there is a 50%
load and the secondary link will be activated. If the router balances the
traffic across the 2 links (don't forget about ip route-cache) then 25Kbps
will travel across both the primary and secondary.
Now suppose the traffic source(s) slows down and there is 18Kbps traffic
presented to the network. This will result in approx 9Kbps on the primary
and secondary. However, since the total traffic rate is still 18kbps, the
10% kick-out (or 10Kbps, since it is 10% of the primary's 100Kbps bandwidth)
the secondary stays active.
Finally, if the traffic source(s) slows its rate to 8Kbps, the primary and
secondary will each carry approximately 4Kbps of traffic. Since the *total*
traffic rate is below the 10% kick-out rate, the secondary link will revert
to 'normal operation' and go into standby mode. The primary link will then
carry the entire 8Kbps.
The BOTTOM LINE is to think of the disable-load value as the traffic rate
which the primary link 'decides' it doesn't need the help of the secondary
any more. This rate is calculated as a percentage of the primary link's
bandwidth. Therefore, the bandwidth of the secondary link does not factor
into the equation at all (BTW, even though it isn't relevant to this issue,
you may want to double check the BW of the BRI interface with a "sh int
bri0"; I think you will find it is 64Kb even though you effectively have
128Kb with the 2 B-channels)
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:30:36 GMT-3