[j-nsp] Traffic shaping on J and SRX

[Michel de Nostredame]

Hi,

Try this config. I am not sure if it could resolve your problem, but you can
give it a try.

==================
interfaces {
    ge-0/0/1 {
        description "## 100m ##";
        unit 0 {
            bandwidth 100m;
            family inet {
                address xx.xx.xx.xx/30;
            }
        }
    }
}
class-of-service {
    interfaces {
        ge-0/0/1 {
            unit 0 {
                shaping-rate 100m;
            }
        }
    }
}
=============================

--
Michel~



On Thu, Jun 3, 2010 at 3:09 PM, Dale Shaw
<dale.shaw+j-nsp at gmail.com<dale.shaw%2Bj-nsp at gmail.com>
> wrote:

> Hi all,
>
> I need to shape on egress from CE to PE to match the service
> provider's 'PVC' speed. The physical interface is gigabit Ethernet. We
> have a variety of J and SRX boxes running JUNOS 10.0R3.
>
> We have the following access/service speed combos (all on J/SRX
> ge-x/x/x or SRX xe-x/x/x): 100m/10m, 1g/50m, 1g/100m, 1g/150m,
> 1g/200m, 10g/1g.
>
> I have a JTAC case running on this but I haven't been making good progress.
>
> The carrier's EoSDH access network throws away traffic when bursts
> fill up the relatively small buffers (~130K) in their Alcatel
> equipment. I need to tune the shaper to control/buffer bursts.
>
> Through the JTAC case (which I have just asked to be escalated), I
> have NOT been able to determine:
>
> 1. how the burst size is calculated when (for example) 'shaping-rate
> 150m' is used without any other configuration on a logical unit.
> 2. how, if at all, the burst size can be tuned. I need to ensure I
> drive the burst size (in bytes) down to <=130K for all of the
> access/speed combinations above
>
> How is everyone doing this in the real world?
>
> Lastly, through my investigations with Juniper folks outside the JTAC,
> I have become aware of two things:
>
> - The default burst size on J-series cannot be configured and
> (whatever it is,) is too high
> - PR/511498. "The shaper burst-size was calculated based on
> interface-bandwidth instead of the shaping-rate". Fixed in
> 10.0-20100326.0 daily SR or later, apparently. No more info.
>
> I hate to draw the comparison but detailed information about how the
> shaper works and can be tuned is readily available in vendor C land
> :-/
>
> Cheers,
> Dale
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>