[j-nsp] Using QoS to ensure BGP bandwidth

Rafał Szarecki rszarecki at gmail.com
Mon Apr 23 16:14:42 EDT 2007


Hi,

   1. some traffic originated on RE is placed in Q3 while other in Q0.
      This can't be changed.
   2. Do you try to apply RED (drop profile)on Q0? This can help, by
      signaling to havy speaker to reduce TCP window.
   3. If RED do not help, then you can try put best effort traffic into
      Q2. This should be easy, however you have to write firewall filter
      on egress. Please note: RE originated packet will be not visable
      for this filer, but all traffic going trough router will be . Guy
      (Hi Guy) wos propose somthing similar but network wide. I believe,
      if this is only problem of this type in your network, better make
      small local fix.

Rafał Szarecki

Jason J. W. Williams napisał(a):
> Hi Guys,
>
> I was hoping to use an egress filter to make sure BGP traffic got top
> priority. The issue that occurred is that we accepted a flood of large
> messages from Hotmail, which was alright on the inbound because the
> Internet acted as speed buffer. However, when we went to deliver them
> it saturated our line outbound going from our LAN to WAN connection,
> which interrupted the BGP healthchecks. We've put some outbound
> bandwidth restrictions on our delivery servers, however I was hoping
> to also put BGP into a higher priority queue than all other traffic to
> ensure the connection stayed up. Thank you in advance for your help.
>
> -J
>
> On 4/23/07, Guy Davies <aguydavies at gmail.com> wrote:
>   
>> Hi All,
>>
>> It's worth pointing out that there is no implicit link between Q0 and
>> a particular scheduler or behaviour.  If you want to use Q0 for
>> control traffic, then do that.  Just make sure that all non-control
>> traffic gets put into a different queue.  That's not particularly
>> difficult if you put a MF classifier at all the ingress points to your
>> network and default all traffic into Q3, for example.
>>
>> Rgds,
>>
>> Guy
>>
>> On 22/04/07, Thomas Mangin <thomas.mangin at exa-networks.co.uk> wrote:
>>     
>>> Alex wrote:
>>>       
>>>> Jason,
>>>> Locally originated BGP traffic is classified into Q0 by default (except BGP
>>>> retransmissions)
>>>> http://www.juniper.net/techpubs/software/junos/junos82/swconfig82-cos/html/cos-hardware4.html#1197876
>>>> --and AFAIK, it is impossible to override this behaviour. So, in an essence,
>>>> you can guarantee bandwidth to
>>>> BGP traffic _through_ the router but not to locally-originated BGP traffic.
>>>>
>>>>         
>>> Never tried, just a mad idea.
>>>
>>> If using trunking to switches : create two vlan, one for control one
>>> traffic. Use the control vlan for BGP traffic but make the next hop the
>>> IP of the traffic interface ? This way you can limit the bandwidth on
>>> the customer vlan and be sure that some is left on the "control" one. No
>>> idea if you can apply the same idea to your IGP as well.
>>>
>>> Thomas
>>> PS: I do not pretend it is a sane idea neither.
>>> _______________________________________________
>>> juniper-nsp mailing list juniper-nsp at puck.nether.net
>>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>>>
>>>       
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
>   


More information about the juniper-nsp mailing list