[j-nsp] Shaping per logical VLAN interface

[Shiva S Shankar]

I understand that "shaping-rate" refers to PIR and "guaranteed-rate" refers to CIR. Correct me if I'm wrong.

Also, if your traffic is bursty in nature you may need to consider the burst-size in bytes what the telco is doing. Also, if you are using DPCs in MX you need to consider that they use the l3 frame data part (without L2 headers) for shaping. Whereas the latest MPC (trio chipset) uses whole L2 frame (including headers IFG etc). The latest being the actual traffic on the wire, would most probably used by Telco, if you are getting a sub-rate cct on a Gig link.

You probably need to use the following command:

set chassis fpc x pic x traffic-manager egress-shaping-overhead x

Cheers



________________________________
 From: Chris Kawchuk <juniperdude at gmail.com>
To: Joao Kluck <gkluck at gmail.com> 
Cc: juniper-nsp at puck.nether.net 
Sent: Tuesday, March 20, 2012 12:11 AM
Subject: Re: [j-nsp] Shaping per logical VLAN interface
 
If the access is a full 1Gig (to the lease provider) and all you want to do is shape each VLAN to 100 Mbit, then do this:

interfaces {
    ge-0/0/0 {
        per-unit-scheduler;
        unit 100 {
           ....vlan/customer specific stuff goes here....
        }
        unit 200 {
           ....vlan/customer specific stuff goes here....
        }
    }
}

class-of-service {
    interfaces {
        ge-0/0/0 {
            unit 100 {
                scheduler-map MyQoS;
                shaping-rate 100m;
            }
            unit 200 {
                scheduler-map MyQoS;
                shaping-rate 100m;
            }
        }
    }
}


If you want to also shape the entire Gig port to an arbitrary throughput (say 500m), then you need to do something like the following:


interfaces {
    ge-0/0/0 {
        hierarchical-scheduler;
    }
}

class-of-service {
    interfaces {
        ge-0/0/0 {
        scheduler-map MyQoS;
        shaping-rate 500m;
            unit 100 {
                output-traffic-control-profile 100m-shaping;
            }
            unit 200 {
                output-traffic-control-profile 100m-shaping;
            }
        }
    }
    traffic-control-profiles {
        100m-shaping {
            scheduler-map MyQoS;  
            shaping-rate 100m;                  
         } 
    }
}

Note: ... I'm writing this from memory/pseudo-code... so you may need to scrub this a bit.

I think you need the "Q" version of the cards in order to do this per VLAN or hierarchical tho. The non-Q cards I believe are only per-port shapers (not capable of per-VLAN); but someone correct me if I'm wrong here...

Hope this helps...!

- CK.



On 2012-03-20, at 10:37 AM, Joao Kluck wrote:

> Dear Community,
> 
> 
> 
> We are analyzing a scenario where we have one MX in a Hub location
> connecting remote sites through a 3rd part leased line provider.
> 
> 
> 
> The MX is connected to the 3rd part provider with 1Gbps physical interface
> with trunked VLAN logical interface.
> 
> The E-lines leased lines connecting MX hub to the remote sites provide
> 100Mbps (CIR=PIR).
> 
> 
> 
> There are 4 different class of service in the internal network and the
> aggregated traffic needs to be shaped at 100Mbps in MX egress interface
> per-destination (i.e VLAN) in order to conform the Leased line SLA provider.
> 
> 
> 
> How it the simplest way to implement this?
> 
> 
> 
> Do we need to implement a kind of HQoS (4x CoS per shaped-VLAN)?
> 
> We intend to use non-Q/EQ MPC.
> 
> 
> Thanks.
> Rgs,
> 
> GK
> _______________________________________________
> 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