[c-nsp] ME3600X Output Drops
Waris Sagheer (waris)
waris at cisco.com
Sat Sep 15 08:20:42 EDT 2012
Hi Ivan,
The policy should work on EFP regardless of bridge or xconnect.
Which image are you using?
Regards,
Waris
-----Original Message-----
From: Ivan [mailto:cisco-nsp at itpro.co.nz]
Sent: Wednesday, September 12, 2012 7:30 PM
To: Waris Sagheer (waris)
Cc: Ivan; cisco-nsp
Subject: RE: [c-nsp] ME3600X Output Drops
Hi Warris,
Thanks again for your answers. The requirement to match at vlan level is because I understood from your examples to apply a policy at the EVC level this was required "<<<<< Using vlan level confirms it is the second level and the child policy is the third level"
I have found that a policy such as the following seems to work and handle both tagged and untagged traffic and can be applied to a EVC. (I wish to apply the same policy to all EVCs regardless of whether the traffic is tagged or not unless customisation is required)
class-map match-any vlan
match vlan 1-4094
policy-map EVC-LEAF
class class-default
queue-limit 491520 bytes
policy-map EVC-VLAN
class vlan
service-policy EVC-LEAF
class class-default
service-policy EVC-LEAF
################
Separate to this, and partly why I was confused, is that I have found the polices applied to the EVC seem to work when bridging but not for a xconnect. Can you confirm this is expected behavior? I would prefer to avoid having to bridge and xconnect on the vlan interface.
interface GigabitEthernet0/5
switchport trunk allowed vlan none
switchport mode trunk
service instance 60 ethernet
encapsulation dot1q 60
service-policy output EVC-VLAN
bridge-domain 60 <--------------- bridged
service instance 70 ethernet
encapsulation dot1q 70
service-policy output EVC-VLAN
xconnect 1.2.3.4 70 encapsulation mpls <--------------- xconnect
Switch#show policy-map interface gi0/5 service instance 60
GigabitEthernet0/5: EFP 60
Service-policy output: EVC-VLAN
Class-map: vlan (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps
Match: vlan 1-4094
Service-policy : EVC-LEAF
Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: any
Queue-limit 491520 bytes
Queue-limit current-queue-depth 0 bytes
Output Queue:
Tail Packets Drop: 0
Tail Bytes Drop: 0
Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: any
Service-policy : EVC-LEAF
Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: any
Queue-limit 491520 bytes <----------------------------queue-limit
Queue-limit current-queue-depth 0 bytes
Output Queue:
Tail Packets Drop: 0
Tail Bytes Drop: 0
Switch#show policy-map interface gi0/5 service instance 70
GigabitEthernet0/5: EFP 70
Service-policy output: EVC-VLAN
Class-map: vlan (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps
Match: vlan 1-4094
Service-policy : EVC-LEAF
Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: any
Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: any
Service-policy : EVC-LEAF
Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: any
> Hi Ivan,
> There is no difference in terms of queue depth in case of policy at
> the port level vs policy at the EVC from hardware programming perspective.
> Port level policy would consume less queues as compare to queues per EVC.
> Policy at EVC has separate queues per service instance. However policy
> at the port level, queues are shared by the all the service instances
> under that port.
> Selection of the policy depends on the requirement. If there are
> separate customers on the EVC then per EVC policy should be used.
> For you last question, shouldn't untagged means no tag then why is
> there a requirement to match on the VLAN level? What kind of traffic
> is there in case of "default"? Is it tagged or untagged?
> If tagged, is there a range?
>
> Regards,
> Waris
>
>
More information about the cisco-nsp
mailing list