01-02-2011 05:54 AM
I have the following policy in a L2 ethernet trunk interface :
policy-map policy-1
class cos-1
priority 10
class cos-2
bandwidth percent 20
class cos-3
bandwidth percent 40
class class-default
bandwidth percent 30
interface GigabitEthernet0/4
switchport mode trunk
switchport trunk allowed vlan 20,30,40
service-policy output policy-1
I want to translate it for an es20 service-instance mode interface like this :
interface GigabitEthernet3/0/13
mls qos trust cos
service instance 1 ethernet
encapsulation dot1q 20
rewrite ingress tag pop 1 symmetric
bridge-domain 20
!
service instance 2 ethernet
encapsulation dot1q 30
rewrite ingress tag pop 1 symmetric
bridge-domain 30
!
service instance 5 ethernet
encapsulation dot1q 40
rewrite ingress tag pop 1 symmetric
bridge-domain 40
!
The problem is that i cant apply a policy on the main interface if i have configure service instances.
How can i share the bandwidth between service instances?
01-02-2011 06:13 AM
If ethernet frames of vlan 20 have cos value 1,
of vlan 30 have cos value 2,
of vlan 40 have cos value 3
I want to share the bandwidth of the main interface to the service instances :
service instance 1 bandwidth 10% with priority,
service instance 2 bandwidth 20% ,
service instance 3 bandwidth 40% ,
service instance 4 bandwidth 30%
01-19-2011 08:51 AM
Hi George,
ES20 doesn't support qos policy on main interface when service instances or sub-interfaces are configured.
The policy has to be directly applied on the service-instance or sub-interface.
In other words service policies on L3 interfaces and L2 service instances are mutual exclusive.
There is no plan to make this available in future releases.
Maybe you can configure a similar policy and apply it to each service instance (Ok, this will not get the same result).
something like the following applied to each service instance:
policy-map policy
class cos-1
police 256000 *
priority
class cos-2
bandwidth percent 20
class cos-3
bandwidth percent 40
class class-default
bandwidth percent 30
* LLQ is supported along with policing only.
Or you need to make the Es20 port a switchport as it was on the other box/card even though I would keep the EVC config which allows you
to configure more staff (beside this type of QoS )
regards,
Riccardo
01-20-2011 03:19 PM
little addendum; if you get an ES+ card you might configure service groups QoS which might get your goal.
Riccardo
01-23-2011 04:22 AM
There is a feature on the SRE version that group the service instances, the evc-group :
Thank you for your reply Riccardo!
01-24-2011 06:28 AM
indeed on SRE we have service-groups (the feature I mentioned) for Es20 too.
I tried it and it works fine.
Riccardo
09-11-2012 12:43 PM
Hi,
It can also be acheived without using service groups as follows:
policy-map child-vlan10
class class-default
priority
police 10000000
queue-limit
!
policy-map parent-vlan10
class class-default
shape average 990000000
bandwidth remaining ratio 10
service-policy child-vlan10
!
policy-map child-vlan20
class class-default
bandwdith percent 20
queue-limit
!
policy-map parent-vlan20
class class-default
shape average 990000000
bandwidth remaining ratio 20
service-policy child-vlan20
!
policy-map child-vlan40
class class-default
bandwidth percent 40
queue-limit
!
policy-map parent-vlan40
class class-default
shape average 990000000
bandwidth remaining ratio 40
service-policy child-vlan40
!
policy-map child-vlan30
class class-default
bandwidth percent 30
queue-limit
!
policy-map parent-vlan30
class class-default
shape average 990000000
bandwidth remaining ratio 30
service-policy child-vlan30
!
interface gigx/x/x
service instance 1 ethernet
service-policy output parent-vlan10
service instance 2 ethernet
service-policy output parent-vlan20
service instance 3 ethernet
service-policy output parent-vlan40
service instance 4 ethernet
service-policy output parent-vlan30
!
This will give you more flexibility in future. Specially if you want to clasify more COS values per service instance which will only require to add extra class in the child policies.
Best regards!
10-29-2012 02:18 PM
I agree with Atif_hafeez
I have also used service groups with success.
However keep in mind usually/hopefully the bandwidth is present in the core network. Let the core just swap labels. they will cary any policies you may have presented on packets
Configure all of your policies at the edge. Ingress/Egress.
Regards
Jude Bryant
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide