RE: Priority command

From: Scott M Vermillion (scott_ccie_list@it-ag.com)
Date: Tue Dec 02 2008 - 02:56:36 ARST


As to the basic question of whether or not the implicit policer is invoked
in the absence of congestion, here is a little test I just threw together
after dinner on my little c877 router here at the house:

c877#sh run | sec class-map
class-map match-any ICMP
 match protocol icmp

c877#sh run | sec policy-map
policy-map TEST_PRIORITY
 class ICMP
  priority 10

c877#sh run int dot0 | inc service-policy
 service-policy output TEST_PRIORITY

(kicks off some large pings around the network that will traverse this
interface and waits for at least 1000 packets)

c877#sh policy-map int dot0
Dot11Radio0

 Service-policy output: TEST_PRIORITY

   Class-map: ICMP (match-any)
     1003 packets, 1446522 bytes
     5 minute offered rate 23000 bps, drop rate 0 bps
     Match: protocol icmp
       1003 packets, 1446522 bytes
       5 minute rate 23000 bps
     Queueing
       Strict Priority
       Output Queue: Conversation 264
       Bandwidth 10 (kbps) Burst 250 (Bytes)
       (pkts matched/bytes matched) 0/0
       (total drops/bytes drops) 0/0

   Class-map: class-default (match-any)
     6573 packets, 3261356 bytes
     5 minute offered rate 41000 bps, drop rate 0 bps
     Match: any

(clears counters and changes configs as follows)

c877#sh run | sec policy-map
policy-map TEST_PRIORITY
 class ICMP
  priority 10
    police 10000

(kicks off identical large pings around the network that will traverse this
interface and waits for at least another 1000 packets)

c877#sh policy-map int dot0
 Dot11Radio0

  Service-policy output: TEST_PRIORITY

    Class-map: ICMP (match-any)
      1002 packets, 1335683 bytes
      5 minute offered rate 16000 bps, drop rate 8000 bps
      Match: protocol icmp
        1002 packets, 1335683 bytes
        5 minute rate 16000 bps
      Queueing
        Strict Priority
        Output Queue: Conversation 264
        Bandwidth 10 (kbps) Burst 250 (Bytes)
        (pkts matched/bytes matched) 0/0
        (total drops/bytes drops) 0/0
      police:
          cir 10000 bps, bc 1500 bytes
        conformed 383 packets, 493857 bytes; actions:
          transmit
        exceeded 619 packets, 841874 bytes; actions:
          drop
        conformed 7000 bps, exceed 7000 bps

    Class-map: class-default (match-any)
      6920 packets, 2396819 bytes
      5 minute offered rate 7000 bps, drop rate 0 bps
      Match: any

So I think that makes pretty clear that the explicit policer was dropping
traffic that the implicit policer wasn't. As for why the "5 minute offered
rate" was a bit lower in the second scenario, I think that had to do with
the end hosts generating the echo requests waiting for timeouts when packets
were dropped by the explicit policer, thus their overall output was slightly
lower. Some of the other counters just don't exactly add up and I really
have no explanation for that. Just different measurement intervals for
different sections of the output? Regardless, this is fairly compelling
IMHO...

(HOPING LIKE HELL THAT THE FORMATTING DOESN'T GET HOSED!!)

Blogs and organic groups at http://www.ccie.net



This archive was generated by hypermail 2.1.4 : Thu Jan 01 2009 - 12:53:07 ARST