[j-nsp] WAN input prioritization on MX
Huan Pham
drie.huanpham at gmail.com
Mon Oct 15 01:09:19 EDT 2012
Hi Caillin,
I can see your points. You think that it is logical to mark traffic as it
comes to the router, and leave it untouched, as it leaves your router. This
is what I used to think of QoS (as I come from the Cisco world). However, I
need to rethink when getting to know Juniper.
With Juniper way, you can still leave the trusted traffic untouched by
"remarking" to the same EXP, or DSCP scheme, as traffic leave your router.
I mean, we are not stuffed.
I do however see a good point in the Juniper way, which marks traffic as it
LEAVES the router!
If you have a managed CE with one LAN connection (connected to customer
LAN), and two WAN connections going to two carriers with 2 different CoS
schemes. You do need to mark traffic differently to match the ISP ones,
depending on which interface it take to exit your router (i.e. depending on
routing).
If you do mark the traffic as it comes to your router, you are stuffed.
Surely, you can say that, you can still remark your "trusted" traffic as it
leaves your router, but it is double marking (you have to do it twice),
isn't it?
Cheers,
Huan
On Mon, Oct 15, 2012 at 2:55 PM, Caillin Bathern <caillinb at commtelns.com>wrote:
> More to the point I believe the original commenter was talking about
> packet marking, not queuing or classification :)
>
> And here I believe that junos doesn't work well... If you have a link
> that carries both transit and newly injected traffic you are stuffed
> when you try to perform a rewrite to correctly mark ingress node traffic
> but also try to transparently pass through traffic from a trusted source
> using the same FC.
>
> Caillin
>
> -----Original Message-----
> From: juniper-nsp-bounces at puck.nether.net
> [mailto:juniper-nsp-bounces at puck.nether.net] On Behalf Of Doug Hanks
> Sent: Monday, 15 October 2012 2:35 PM
> To: Serge Vautour; Chris Evans; Gustavo Santos
> Cc: juniper-nsp at puck.nether.net
> Subject: Re: [j-nsp] WAN input prioritization on MX
>
> Yes, that's just what I said in so few words :-)
>
> Classification = ingress
> Queuing = egress
>
> From: Serge Vautour
> <sergevautour at yahoo.ca<mailto:sergevautour at yahoo.ca>>
> Reply-To: Serge Vautour <serge at nbnet.nb.ca<mailto:serge at nbnet.nb.ca>>
> Date: Sun, 14 Oct 2012 10:06:37 -0700
> To: dhanks <dhanks at juniper.net<mailto:dhanks at juniper.net>>, Chris Evans
> <chrisccnpspam2 at gmail.com<mailto:chrisccnpspam2 at gmail.com>>, Gustavo
> Santos <gustkiller at gmail.com<mailto:gustkiller at gmail.com>>
> Cc: "juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>"
> <juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>>
> Subject: Re: [j-nsp] WAN input prioritization on MX
>
> Humm. My understand, at least with the command sets I'm use to using, is
> that you do classification on ingress and then queuing and marking on
> egress. When you do classification, the packets are assigned to a
> "Forwarding Class (FC)" inside the box. This makes sure the box gives
> those packets proper treatment inside the box and that the packets get
> assigned to the proper egress interface queue. While the packets exit
> the queue (based on egress schedulers), the packet QoS headers are
> remarked.
>
> Basically, this diagram:
>
> http://www.juniper.net/techpubs/images/g017213.gif
>
> Packets travel through the box based on the outer boxes following the
> solid lines. The dotted lines all point to or from the FC to identify
> how the decision is made.
>
> Serge
>
>
> ________________________________
> From: Doug Hanks <dhanks at juniper.net<mailto:dhanks at juniper.net>>
> To: Chris Evans
> <chrisccnpspam2 at gmail.com<mailto:chrisccnpspam2 at gmail.com>>; Gustavo
> Santos <gustkiller at gmail.com<mailto:gustkiller at gmail.com>>
> Cc: "juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>"
> <juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>>
> Sent: Sunday, October 14, 2012 12:09:53 AM
> Subject: Re: [j-nsp] WAN input prioritization on MX
>
> How is this weird? You can mark on ingress, but the queuing happens on
> the egress interface when it's to be transmitted.
>
>
> On 10/13/12 6:07 AM, "Chris Evans"
> <chrisccnpspam2 at gmail.com<mailto:chrisccnpspam2 at gmail.com>> wrote:
>
> >JUNOS does a weird way of marking packets.. It is done on the egress of
>
> >the box, not on ingress (there is an exception in a few newer modules
> >that can do this). So it is probably working as the other poster
> >mentioned. Make sure you take this methodology into consideration as
> >it can hinder your granularity of CoS with marking vs passing through
> >and you inadvertently remark traffic you didn't mean to.
> >
> >On Sat, Oct 13, 2012 at 8:21 AM, Gustavo Santos
> ><gustkiller at gmail.com<mailto:gustkiller at gmail.com>>wrote:
> >
> >> Doug and Hanks @juniper. I had to left the office and leave
> >>configuration as is. On monday I will update you after verify what
> >>you have pointed,
> >>
> >> What I can tell is that I didn't have made any modification on the
> >>systems default class of service / mapping configuration.
> >>
> >> Thank you!
> >>
> >> Gustavo Santos
> >> Analista de Redes
> >> CCNA , MTCNA , MTCRE, MTCINE, JUNCIA-ER
> >>
> >>
> >>
> >> 2012/10/13 Harry Reynolds
> >> <harry at juniper.net<mailto:harry at juniper.net>>
> >>
> >> > Doug raises some good points.
> >> >
> >> > Also, for testing, perhaps add some counters to the terms to aid in
>
> >> > confirming matches. You may also want to show config | display
> >> > detail/inheritance to see if the prefix list is expanding as you
> >>expect.
> >> >
> >> > Regards
> >> >
> >> >
> >> >
> >> > -----Original Message-----
> >> > From:
> juniper-nsp-bounces at puck.nether.net<mailto:juniper-nsp-bounces at puck.neth
> er.net> [mailto:
> >> > juniper-nsp-bounces at puck.nether.net<mailto:juniper-nsp-bounces at puck
> >> > .nether.net>] On Behalf Of Doug Hanks
> >> > Sent: Friday, October 12, 2012 9:36 PM
> >> > To: Gustavo Santos;
> >> > juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>
> >> > Subject: Re: [j-nsp] WAN input prioritization on MX
> >> >
> >> > I'm sure it's working just fine. Are you checking the egress
> >>interface to
> >> > see if the traffic is being marked and queued properly? A common
> >>mistake
> >> is
> >> > to check the ingress interface queues.
> >> >
> >> >
> >> > If this doesn't work, we would need to see your entire
> >>class-of-service
> >> > configuration.
> >> >
> >> > On 10/12/12 6:04 PM, "Gustavo Santos"
> <gustkiller at gmail.com<mailto:gustkiller at gmail.com>> wrote:
> >> >
> >> > >Hi,
> >> > >
> >> > >I'm new on Juniper class of service / shaping. I'm reading some
> >> > >tech docs from Juniper and a Juniper's MX book, but it's kind
> tricky.
> >> > >Today I get asked to do a pretty simple configuration, but I tried
> >>some
> >> > >settings but none of then worked. Any of you guys can help me with
> >>that?
> >> > >
> >> > >What I want to achieve is pretty (conceptualy speaking) simple. I
> >>have
> >> > >a Gig interface and want to rate limit the interface at 500Mbits ,
> >>mark
> >> > >a destination subnet with expedited forwarding class, mark
> >> > >anything else with best effort. I tried the config below but it's
> not working.
> >> > >The rate-limit works but the prioritization isn't.
> >> > >
> >> > >
> >> > >
> >> > >
> >> > >gustavo at MX5-1> show configuration firewall family inet filter
> >> > >wan-control physical-interface-filter; term high-priority {
> >> > > from {
> >> > > destination-prefix-list {
> >> > > high-priority-dst;
> >> > > }
> >> > > }
> >> > > then {
> >> > > policer limit500;
> >> > > loss-priority low;
> >> > > forwarding-class expedited-forwarding;
> >> > > }
> >> > >}
> >> > >term else {
> >> > > then {
> >> > > policer limit500;
> >> > > loss-priority high;
> >> > > forwarding-class best-effort
> >> > > }
> >> > >
> >> > >
> >> > >( policer limit500)
> >> > >physical-interface-policer;
> >> > >if-exceeding {
> >> > > bandwidth-limit 480m; (set the value lower to check policer
> >> > >working..
> >> > >but it wasn't as desired)
> >> > > burst-size-limit 625k;
> >> > >}
> >> > >then discard;
> >> > >
> >> > >then the filter was applied on the interface family inet filter
> >> > >input wan-control
> >> > >
> >> > >Gustavo Santos
> >> > >Analista de Redes
> >> > >CCNA , MTCNA , MTCRE, MTCINE, JUNCIA-ER
> >> > >_______________________________________________
> >> > >juniper-nsp mailing list
> >> > >juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>
> >> > >https://puck.nether.net/mailman/listinfo/juniper-nsp
> >> > >
> >> >
> >> >
> >> >
> >> > _______________________________________________
> >> > juniper-nsp mailing list
> >> > juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>
> >> > https://puck.nether.net/mailman/listinfo/juniper-nsp
> >> >
> >> >
> >> >
> >> _______________________________________________
> >> juniper-nsp mailing list
> >> juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>
> >> https://puck.nether.net/mailman/listinfo/juniper-nsp
> >>
> >_______________________________________________
> >juniper-nsp mailing list
> >juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>
> >https://puck.nether.net/mailman/listinfo/juniper-nsp
> >
>
>
>
> _______________________________________________
> juniper-nsp mailing list
> juniper-nsp at puck.nether.net<mailto: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
> --
> Message protected by MailGuard: e-mail anti-virus, anti-spam and
> content filtering.http://www.mailguard.com.au/mg
>
>
> _______________________________________________
> 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