Re: Inbound Policier on the 3550 not accurate

From: Pierre-Alex (paguanel@hotmail.com)
Date: Fri Jun 02 2006 - 10:33:31 ART


OK, thanks!
  ----- Original Message -----
  From: Chris Lewis
  To: Pierre-Alex
  Cc: ccielab@groupstudy.com
  Sent: Friday, June 02, 2006 3:03 PM
  Subject: Re: Inbound Policier on the 3550 not accurate

  Hi Pierre,

  There are a number of issues that could give rise to such readings, however
the method of relying on pings and show commands is not accurate in terms of
measuring policer accuracy. The policers measure L3 bandwidth and don't take
L2 headers in to account, hence the only way to measure the accuracy of them
is to use something like a smartbits exteranlly to the router.

  Chris

  On 6/2/06, Pierre-Alex <paguanel@hotmail.com> wrote:
    I have the following problem: when applying a policy inbound on a trunk
port I
    have a discrepency of 40 K. See Verification section.
    I do not have the problem when I am classifying via IP addresses. Has
anyone
    of you observed this? Any fix appart from changing the policing values?

    Diagram:

    Diagram :

    r2 ------trunk -----Switch 1 -----R3
                                |
                              R6

    r6 is in vlan 26
    r3 is in vlan 23

    -----------

    Task:

    R2 is pinging r6 and r3 at line rate with respective precedence of 1 and
5.
    Limit traffic to r6 to 500Kb/s and
    limit traffic to R3 at 1 Megabit/s. Drop the excess.

    ----------------

    Configuration

    mls qos

    !

    interface FastEthernet0/2
    switchport trunk encapsulation dot1q
    switchport mode trunk
    switchport nonegotiate
    service-policy input police-in
    mls qos monitor dscp 8 40

    class-map match-all cs1
    match ip dscp cs1
    class-map match-all cs5
    match ip dscp cs5

    policy-map police-in
    class cs1
       police 500000 93750 exceed-action drop
    class cs5
       police 1000000 187500 exceed-action drop

    FastEthernet0/2
    Ingress
    dscp: incoming no_change classified policed dropped (in bytes)
       8 : 262354422 51393408 0 0 0
       40: 102161400 0 0 0 0
    Others: 494922022 445330872 362713564 0 294402306

    --------------

    Verification

    r3#sh policy-map interface
    Ethernet0/0

    Service-policy input: inbound_measure

       Class-map: all_traffic (match-all)
         557654 packets, 363809819 bytes
         30 second offered rate 1046000 bps
         Match: access-group 1

       Class-map: class-default (match-any)
         0 packets, 0 bytes
         30 second offered rate 0 bps, drop rate 0 bps
         Match: any

    r6#sh policy-map interface
    Ethernet0/0

    Service-policy input: inbound_measure

       Class-map: all_traffic (match-all)
         878931 packets, 449601296 bytes
         30 second offered rate 540000 bps
         Match: access-group 1

       Class-map: class-default (match-any)
         32687 packets, 2457912 bytes
         30 second offered rate 0 bps, drop rate 0 bps
         Match: any

    !NB the policy-map on r3 and r6 do not do any actions. They are there only
to
    measure incoming traffic.
    You can see that the 30 seconds rate is of by about 40K in both cases.

    _______________________________________________________________________
    Subscription information may be found at:
    http://www.groupstudy.com/list/CCIELab.html



This archive was generated by hypermail 2.1.4 : Sat Jul 01 2006 - 07:57:31 ART