upstream1 peering upstream2
\ | /
\ | /
\ | /
juniper m5 ---- c7206vxr
/ | | \
/ | | \
customer customer customer customer
filter upstream1 {
term one {
from interface group 1;
then {
forwarding-class internet;
accept;
}
}
term two {
then accept;
}
}
filter upstream2 {
term one {
from {
dscp af23;
}
then {
forwarding-class internet;
accept;
}
}
term two {
then accept;
}
}
[edit interfaces]
interfaces upstream1 {
unit logical-unit-number {
family inet {
filter {
group 1 ;
input upstream1 ;
}
}
}
}
interfaces m5-to-cisco {
unit logical-unit-number {
family inet {
filter {
input upstream2 ;
}
}
}
}
class-of-service {
forwarding-classes {
queue 2 internet;
}
interfaces {
juniper-to-junipercustomer {
scheduler-map foo;
}
}
scheduler-maps {
foo {
forwarding-class internet scheduler internet;
}
}
schedulers {
internet {
transmit-rate 2m exact; <= if you want the minimum
guarantee 2M, but if no other traffic from other customers or from your
internet peer, you don't need to put "exact" here, so the whole port can
be used for transmitting the international traffic from upstream1 and 2.
}
}
Jane
> -----Original Message-----
> From: Josef Buchsteiner
> Sent: Thursday, April 04, 2002 8:32 AM
> To: Blaz Zupan; Ben Eater
> Cc: juniper-nsp@puck.nether.net
> Subject: RE: [j-nsp] DSCP marking for packets not leaving router
>
>
> At 05:54 PM 4/3/2002, Blaz Zupan wrote:
> > > I think it sounds like you want to use interface-groups.
> Assign the
> > > interface to the upstream provider to an interface group:
> >
> >Hey, that looks like it might be the solution. The only
> problem is, that I
> >would like to also limit packets that are marked with a
> certain DSCP by the
> >Cisco (see the ASCII art of the real situation in the reply
> I sent to Eduard).
> >
> >What would something like this do:
> >
> >filter foo {
> > term one {
> > from interface-group 1;
> > then policer bar;
> > }
> > term two {
> > from dscp af43;
> > then policer bar;
> > }
> > term three {
> > then accept;
> > }
> >}
> >
> >Let's say bar limits packets to 2Mbps. Would the above limit
> packets comming
> >from interface group to 2Mbps and packets comming in with
> DSCP af43 to 2Mbps?
> >Thus, the aggregate usage could be up to 4Mbps?
>
> correct
>
> > (Which is not what I would
> >like).
> >
> >In pseudo-code, what I would like to achieve is something like this:
> >
> >
> >filter foo {
> > term one {
> > from <packet comming from interface-group 1 *or* marked
> with dscp af43>;
> > then policer bar;
> > }
> > term two {
> > then accept;
> > }
> >}
> >
> > >From the options I have, I don't seem to be able to
> achieve the above.
>
> what about the following ?
>
> term 1 {
> from {
> interface-group 1;
> }
> then {
> forwarding-class internet;
> accept;
> }
> }
> term 2 {
> from {
> dscp af23;
> }
> then {
> forwarding-class internet;
> accept;
> }
> }
> term 3 {
> then {
> accept;
> }
> }
> }
>
>
> You assign this filter to the interface going to your internet peer
> and now you define you class-of-service scheduler for this interface
>
>
> <-- keep in mind this is very limited configuration and you
> may want to
> <-- set the proper defined the other queues like queue 3 for
> network-control
> <-- but this is just a simplified way to show how you can accomplish
> <-- this
>
> class-of-service {
> forwarding-classes {
> queue 2 internet;
>
> }
> interfaces {
> so-0/0/3 {
> scheduler-map foo;
> }
> }
> scheduler-maps {
> foo {
> forwarding-class internet scheduler internet;
> }
> }
> schedulers {
> internet {
> transmit-rate 2m exact;
> }
> }
>
>
>
>
>
>
>
>
>
>
This archive was generated by hypermail 2b29 : Mon Aug 05 2002 - 10:42:35 EDT