[f-nsp] VPLS/VLL Redundancy

[Tomasz Szewczyk]

Hi,

I think now I got your point. In my opinion placing the RLs depends on 
what is the service definition for the customer. I think the simplest 
solution is to put RL on access interface, but in case of VPLS-like 
services (many-to-many) it is not so easy to predict traffic amount in 
the core. For example if customer wants to connect 5 sites with 100Mbps, 
I would put input rate-limit and output rate-shaping (both 100Mbps) for 
each access interface. RL should limit traffic entering the core and 
priority/DSCP based rate shaping avoid (or manage) congestion on output 
interface.

Best Regards

Tomek

Lazuardi Nasution pisze:
> Hi Tomasz,
>
> What if there are some another CEs of the same subscriber/company
> which are connected to the same aggregation switch ? I think I must
> assign them into the same VLAN for making them to be the same member
> of VPLS cloud ? If so, the traffic from a CE to the another CE will
> not traverse to VPLS Router. In that case, VLAN based RL on VPLS
> Router side cannot be used.
>
> Best regards,
>
> On Fri, Oct 17, 2008 at 4:13 PM, Tomasz Szewczyk <tomeks at man.poznan.pl> wrote:
>   
>> Hi,
>> I think you can use vlan baser RL but in this case I would suggest to run
>> STP on access switches and for some VPLS instances (it is possible on 3.8 or
>> 3.9 code on XMR as Topology Group). This would give you better control on
>> limits applied for the customer, because you can setup primary and secondary
>> active port. I mean something like on attached picture. You can apply the
>> same RL on ports A and B. The port B should be blocked by STP, hence all
>> customer traffic will be limited by policer on port A. In case of failure
>> port B becomes active and all customer traffic will be limited by policer
>> applied to port B.
>> But this relies on STP - so in my opinion it is the last thing you should do
>> ;-) Additional problem is L2 VLAN for STP control. The main disadvantage is
>> that VPLS can not be used for STP control, so you have to connect XMRs via
>> L2 (in our network we don't have such connections).
>>
>> Regards
>>
>> Tomek
>>
>> Lazuardi Nasution pisze:
>>     
>>> Hi Tomasz,
>>>
>>> The configuration is like attached picture. Each aggregation switch is
>>> connected to two VPLS Routers as a ring. I use two  VPLS Router for
>>> redundancy. Each subscriber device will be connected to one or two
>>> aggregation switches side. So, I think the port based rate limitting
>>> must be done on aggregation switches. I have no idea if I could use
>>> VLAN based rate limitting on the VPLS router side since there are some
>>> subsribers which have more than one node connected to the same
>>> aggregation switch.
>>>
>>> Best regards.
>>>
>>>
>>> On Fri, Oct 17, 2008 at 1:48 PM, Tomasz Szewczyk <tomeks at man.poznan.pl>
>>> wrote:
>>>
>>>       
>>>> Hi,
>>>>
>>>>
>>>>         
>>>>> 1. Anyone has experience with the VPLS/VLL redundancy where each
>>>>> subscriber node is connected to diffrent PE Router (NetIron MLX/XMR) ?
>>>>> I need some configuration brief about that. The whitepapers are very
>>>>> helpful too.
>>>>>
>>>>>
>>>>>           
>>>> We used transparent BPDU flooding over VPLS instance, so it was possible
>>>> to
>>>> run STP on L2 devices connected to XMRs (bot not on XMRs)
>>>> It is possible to enable it on interface:
>>>> no vpls-bpdu-block
>>>> However I used this only in case when L2 devices were under our
>>>> administration. Running this option on customer's interface is a little
>>>> bit
>>>> risky because in case of L2 loop created by customer, the control plane
>>>> of
>>>> XMRs can be affected. However there is cpu-protection option, but I think
>>>> it
>>>> will not protect you against "loop" MAC learning.
>>>>
>>>>         
>>>>> 2. Is it possible to map different VLANs into the same VPLS cloud ?
>>>>>
>>>>>
>>>>>           
>>>> Yes - you just need to type:
>>>> vlan xxx
>>>> tag/untag e8/1
>>>> vlan yyy
>>>> tag/untag e8/2
>>>>
>>>> Or you just setup different VLANs on remote VPLS endpoints. It works :-)
>>>>
>>>>         
>>>>> 3. I need some suggestion of FastIron Edge products which have port
>>>>> based incoming and outgoing rate limitting with 64kbps granularity. I
>>>>> will combine it on my design with NetIron MLX/XMR for making VPLS/VLL
>>>>> provider.
>>>>>
>>>>>
>>>>>           
>>>> Unfortunately I don't have a lot of experiences with FastIron series, but
>>>> last times I successfully tested rate-shaping for VPLS based on DSCP in
>>>> customer's IP packets. It works fine too :-)
>>>>
>>>> Regards
>>>>
>>>> Tomek
>>>>
>>>>
>>>>
>>>> ------------------------------------------------------------------------
>>>>
>>>>