[j-nsp] Queue management, follow-up on configuring CoS on ATM2 interfaces.

Roberts, Michael J. (IATS) RobertsMJ at missouri.edu
Fri Jun 4 15:46:17 EDT 2004


Since my last post, I have successfully implemented QoS on the ATM and
GigE interfaces on our M20. I followed the advice of several people (My
SE helped a lot), and ripped off some configuration ideas from
previously posted examples. 

 

My schedulers are configured as the following:

 

schedulers {

    ef-scheduler {

        transmit-rate percent 5;

        buffer-size percent 5;

        priority high;

    }

    be-scheduler {

        transmit-rate percent 90;

        buffer-size percent 90;

        priority low;

    }

    nc-scheduler {

        transmit-rate percent 5;

        buffer-size percent 5;

        priority high;

    }

}

 

 

ATM:

scheduler-maps {

        ATMToSSchedulerMap {

            forwarding-class expedited-forwarding {

                priority high;

                transmit-weight percent 5;

            }

            forwarding-class best-effort {

                priority low;

                transmit-weight percent 90;

            }

            forwarding-class network-control {

                priority low;

                transmit-weight percent 5;

            }

        }

 

 

I am only concerned with three queues at the moment and I do not have
WRED or RED configured on any of them.  My question is the following:  I
have discovered that my traffic being transmitted through the expedited
forwarding queue is able to transmit at rates much higher than 5% of the
link speed.  This is not a problem in the lab, where the links do not
experience any congestion, but I am curious to know what the 'rules' are
for queues bursting into allocated space of the other queues and how
this behavior changes when the link becomes congested.  I would to get a
better grip on the expected behavior under heavy load.  Any insight
would be great.  THx.

 

-mike



More information about the juniper-nsp mailing list