[c-nsp] FR QOS

Oliver Boehmer (oboehmer) oboehmer at cisco.com
Sat Mar 5 11:32:22 EST 2005


> As of now, I am testing on Cisco 3845 Series. After I gave the Shape
> command, it has worked. I am right now testing in the lab and
> scenario is like this  
> 
> Cisco 7507 (PE)<---------E1 Link (2 PVC's)back to back------->Cisco
> 3845(CE) 

Ok. Is this a Multi-VRF setup on the CE?
 
> Couples of queries
> 
> 1) Does the "bandwidth percent xx" and "shape average percent 50"
> command which we are applying in the classes mean the same? 

well, no. "bandwidth" reserves a certain BW for the respective class,
"shape" limits the traffic to the configured BW. You need the shaper at
the parent level in order to queue within the child policy. I must say I
don't really know if this will really do what you want on the CE, maybe
you want to use a different approach and define a common policy-map
(non-hierarchical) for both customers where you match on both precedence
(or other criteria) as well as the DLCI (via "class-map match-all
xxxx").

I've never actually used QoS in this environment, so not realy sure what
the best way is. Maybe someone else on the list has this running?
 
> 2) Can you little elaborate the statement "both PVCs over the
> Provider network should be configured to be able to peak to the full
> Line speed. If PVCs were limited at the provider network each to a
> Specific bandwidth, this QoS approach won't work."

Well, assume you deploy this in a "real" FR environment (with FR
switches) and you get two 512k CIR PVCs on an E1/T1, you will obviously
not be able to use more than 512k per PVC, even if the other PVC carries
no traffic. This isn't an issue in a back-to-back config like yours
where one PVC will be able to burst up the physical line speed..

	oli
 
> 
> -----Original Message-----
> From: Oliver Boehmer (oboehmer) [mailto:oboehmer at cisco.com]
> Sent: Saturday, March 05, 2005 3:22 PM
> To: Akella Vardhana Srikant; cisco-nsp at puck.nether.net
> Subject: RE: [c-nsp] FR QOS
> 
> 
>> 
>> I have done the same config but I am getting the below error message
>> . Could you please let me know in case any modifications are
>> required in the config. 
>> 
>> chester(config)#int s2/0
>> chester(config-if)#service-policy output Out-s4/1/0
>> CBWFQ : Hierarchy supported only if shaping is configured in this
>> class chester(config-if)#
> 
> 
> This is what I mentioned, some platforms require you to apply shaping.
> Which platform do you have? Can you add a "shape average <x>" in your
> "class dlci-401"?
> 
> What is the exact environment? Is this back-to-back FR connection?
> 
> 	oli
> 
> 
>> policy-map Out-s4/1/0
>>  class dlci-401
>>   bandwidth percent 50
>>   service-policy customer-1
>>  class dlci-402
>> 
>>   bandwidth percent 25
>> 
>>   service-policy customer-2
>> 
>> 
>> 
>> chester#sh run int s2/0
>> 
>> Building configuration...
>> 
>> 
>> 
>> Current configuration : 136 bytes
>> 
>> !
>> 
>> interface Serial2/0
>> 
>>  description To Stanster
>> 
>>  no ip address
>> 
>>  encapsulation frame-relay IETF
>> 
>>  map-group test
>> 
>>  dsu bandwidth 44210
>> 
>> end
>> 
>> 
>> 
>> Regards,
>> 
>> Srikant
>> 
>> 
>> 
>> -----Original Message-----
>> From: Oliver Boehmer (oboehmer) [mailto:oboehmer at cisco.com]
>> Sent: Saturday, March 05, 2005 3:02 PM
>> To: Akella Vardhana Srikant; cisco-nsp at puck.nether.net
>> Subject: RE: [c-nsp] FR QOS
>> 
>> 
>> 
>> 
>> 
>>> I have some queries
>> 
>>> 
>> 
>>> policy-map Out-s4/1/0
>> 
>>>  class dlci-401
>> 
>>>   bandwidth percent 50
>> 
>>>   service-policy customer-1
>> 
>>>  class dlci-402
>> 
>>>   bandwidth percent 25
>> 
>>>   service-policy customer-2
>> 
>>> 
>> 
>>> 
>> 
>>> In the above statements, you are hardcoding a bandwidth percent
>>> value 
>> 
>>> under of the class Like "dlci-401" and "dlci-402" . After you define
>> 
>>> this , does the Router allow us to share bandwidth between VPNs.
>> 
>> 
>> 
>> Sure, this is the reason why you need to apply the policy at the
>> 
>> physical interface. If VPN-1 doesn't use "his" bandwidth, VPN-2 can
>> use 
>> 
>> it.
>> 
>> 
>> 
>>       oli
>> 
>> 
>> 
>>> 
>> 
>>> -----Original Message-----
>> 
>>> From: Oliver Boehmer (oboehmer) [mailto:oboehmer at cisco.com]
>> 
>>> Sent: Saturday, March 05, 2005 2:51 PM
>> 
>>> To: Akella Vardhana Srikant; cisco-nsp at puck.nether.net
>> 
>>> Subject: RE: [c-nsp] FR QOS
>> 
>>> 
>> 
>>> 
>> 
>>>> I need some help with respect to the FR QOS. In our scenario, we
>>>> have
>> 
>>>> 2 PVC's configured under single E1 Link and PVC belongs to
>>>> respective
>> 
>>>> customer. The respective PVC has again 4 different types of traffic
>> 
>>>> such as Voice, FTP, etc.
>> 
>>>> 
>> 
>>>> The Demand is that Total E1 BW should be utilized among to the 2
>> 
>>>> VPN's and again among the respective classes of traffic such as
>> 
>>>> Voice, FTP, etc. The QOS config should be defined in such a way
>>>> that 
>> 
>>>> in case PVC-1 is down then PVC-2 should utilize all the E1 BW and
>> 
>>>> then sub divide among all the 4 classes.
>> 
>>>> 
>> 
>>>> Similarly it should be other way round also.
>> 
>>>> 
>> 
>>>> Can some body tell me what sort of QOS config has to deployed for
>> 
>>>> achieving this ?
>> 
>>> 
>> 
>>> You need some form of hierarchical CBWFQ using the "match fr-dlci"
>> 
>>> class-map classification in the parent and your customer-specific
>> 
>>> classes at the child. you would then apply the policy-map at the
>> 
>>> physical FR interface, not at the customer-specific sub-interfaces.
>> 
>>> Please note that some platforms (for example 3600) require you to
>> 
>>> shape
>> 
>>> at the parent.
>> 
>>> 
>> 
>>> Here's an example:
>> 
>>> 
>> 
>>> class-map match-all platin
>> 
>>>  match ip precedence 5
>> 
>>> class-map match-all gold
>> 
>>>  match ip precedence 3
>> 
>>> class-map match-all silber
>> 
>>>  match ip precedence 2
>> 
>>> !
>> 
>>> class-map match-all dlci-402
>> 
>>>  match fr-dlci 402
>> 
>>> class-map match-all dlci-401
>> 
>>>  match fr-dlci 401
>> 
>>> !
>> 
>>> !
>> 
>>> policy-map customer-1
>> 
>>>  class platin
>> 
>>>   priority 64
>> 
>>>  class gold
>> 
>>>   bandwidth 64
>> 
>>>  class silber
>> 
>>>   bandwidth 64
>> 
>>>  class class-default
>> 
>>>   random-detect
>> 
>>>   bandwidth 64
>> 
>>> !
>> 
>>> policy-map customer-2
>> 
>>>  ....
>> 
>>> !
>> 
>>> policy-map Out-s4/1/0
>> 
>>>  class dlci-401
>> 
>>>   bandwidth percent 50
>> 
>>>   service-policy customer-1
>> 
>>>  class dlci-402
>> 
>>>   bandwidth percent 25
>> 
>>>   service-policy customer-2
>> 
>>> !
>> 
>>> interface Serial4/1/0
>> 
>>>  bandwidth 512000
>> 
>>>  no ip address
>> 
>>>  no ip directed-broadcast
>> 
>>>  encapsulation frame-relay
>> 
>>>  load-interval 30
>> 
>>>  frame-relay intf-type dce
>> 
>>>  service-policy output Out-s4/1/0
>> 
>>> !
>> 
>>> int Serial4/1/0.1
>> 
>>>  ip address ...
>> 
>>>  frame-relay interface-dlci 401
>> 
>>> int Serial4/1/0.2
>> 
>>>  ip address ...
>> 
>>>  frame-relay interface-dlci 402
>> 
>>> 
>> 
>>>     oli



More information about the cisco-nsp mailing list