[c-nsp] QoS Question

[Rodney Dunn]

The best way to do it is like this.

I'm assuming your congestion is on the WAN T1's.

On the ingress LAN links at the spokes you define
a service policy to match and mark your traffic with
ip precedence or DSCP.  For any traffic you don't
specifically match you set it to 0 for the default class.
That is your marking policy.

Then, on the WAN links you simply define a policy
egress to do the appropriate queueing and for the
match classes on that side you simply use the
marking from the inbound policy. 


ie:


interface fe 0
service-policy input marking

interface serial 0
service-policy output queueing

!classes to match traffic on ingress to set prec accordingly

class-map match-all markimportant
 match access-group 101
class-map match-all lessimportant
 match access-group 102

!now define the policy that will do the sets on ingress

policy-map marking
 class markimportant
  set ip precedence 5
 class lessimportant
  set ip precedence 4
 class class-default
  set ip precedence 0


!egress classes to match for queueing

class-map match-all queueing4important
 match ip precedence 5
class-map match-all queueing4lessimportant
 match ip precedence 4

!policy to do the queueing egress on the wan links

policy-map egressqueueing
 class queueing4important
  bandwidth X or bandwidth percent Y (whichever you want to do)
 class queueing4lessimportant
  bandwidth X or bandwidth percent Y (whichever you want to do)
 class class-default
  fair-queue
  random-detect

This QOS policy says that during congestion the class you define
with bandwidth X is guaranteed to get that amount of total
bandwidth during congested conditions.

All the traffic that doesn't match one of the BW classes get's
best effort service with the BW left over for the default class.

Your ACL's ingress will determine what egress QOS class the
packets get in.

Rodney


On Thu, Sep 23, 2004 at 08:27:43AM -0400, Chapman, Matthew wrote:
a
> Hello,
> 
> I have a basic QoS question.  I have a hub and spoke topology, all T1
> connections.  Some major applications are being done via http at the
> hub.  We have a class B that we use mostly for servers, and no remote
> locations use any of the class B.  We have a few critical websites that
> are on the internet just beyond our hub that the remote locations need
> access to above all other sites.
> 
> I would like to make a QoS policy ingress on the router at the remote
> location that gives a high priority to anything in the class B (mostly
> to insure Microsoft AD does not break etc and our new web apps work) and
> then give a lower priority to a few select websites on the internet and
> then everything else fall into "if there is enough bandwidth left"
> category.
> 
> Any suggestions on this?  Is this the right approach?  
> 
> Thanks,
> Matt Chapman
>  
> --
> Matthew Chapman
> Network Engineer
> "Hard work, sacrifice and focus will never show up in tests." 
>          - Lance Armstrong 
> 
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/