From: gladston@br.ibm.com
Date: Wed Nov 17 2004 - 09:29:35 GMT-3
As I understand it, the first way should not be used on the second method.
As I remember it, the idea of the formula is to reach a result similar to WRED, instead of droping many packets on bursts an having synchronization problem when policing.
"normal burst = configured rate * (1 byte)/(8 bits) * 1.5 seconds
extended burst = 2 * normal burst"
Taking in account that the second bucket has its own rate filling the bucket, "Be" would represent the depth of the second bucket, so it would be wise to set a value that represents the burst when sending data on peak rate. For example, say the application normally sends 512kbps and the burst is 8000 bytes; and ocassionaly, the traffic is sent at peaks of 1Mbps and burst of 16000 bytes. The command would be "police cir 512000 bc 8000 pir 1000000 be 16000..."
=======
quoted
bc = CIR * tc/1000
be = ( AR - CIR ) * Tc/1000
========
Maybe the formula should be:
bc = CIR * tc/1000
be = PIR * Tc/1000
Any other comments?
This archive was generated by hypermail 2.1.4 : Thu Dec 02 2004 - 06:57:46 GMT-3