Re: [j-nsp] QFX5100 Buffer allocation for burst be traffic

[Nitzan Tzelniker]

We are doing using this config and it reduce the amount of drops
IIRC I did  the change without any issue


Nitzan

On Tue, May 30, 2017 at 2:26 PM, Chen Jiang  wrote:

> Hi! Experts
>
> Sorry for disturbing, we are using QFX5100 virtual-chassis for
> distrubiuting storage traffic, the average traffic volume is 2-3 Gbps but
> there is burst traffic will exceed the interface bandwidth (we see there
> are tailed drop counter keep increasing in best-effort queue), so we want
> to reallocate maximum switch buffer for  egress be traffic, from the JUNOS
> document it seems there is a knob to do this:
>
> lab@qfx5100> show configuration class-of-service shared-buffer
> ingress {
> percent 100;
> buffer-partition lossless {
> percent 5;
> }
> buffer-partition lossy {
> percent 90;
> }
> buffer-partition lossless-headroom {
> percent 5;
> }
> }
> egress {
> percent 100;
> buffer-partition lossless {
> percent 5;
> }
> buffer-partition lossy {
> percent 90;
> }
> buffer-partition multicast {
> percent 5;
> }
> }
>
> Is there any experience or any best practice to do this? Thanks for your
> help!
>
>
> --
> BR!
>
>
>
>James Chen
> ___
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
___
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

[j-nsp] QFX5100 Buffer allocation for burst be traffic

[Chen Jiang]

Hi! Experts

Sorry for disturbing, we are using QFX5100 virtual-chassis for
distrubiuting storage traffic, the average traffic volume is 2-3 Gbps but
there is burst traffic will exceed the interface bandwidth (we see there
are tailed drop counter keep increasing in best-effort queue), so we want
to reallocate maximum switch buffer for  egress be traffic, from the JUNOS
document it seems there is a knob to do this:

lab@qfx5100> show configuration class-of-service shared-buffer
ingress {
percent 100;
buffer-partition lossless {
percent 5;
}
buffer-partition lossy {
percent 90;
}
buffer-partition lossless-headroom {
percent 5;
}
}
egress {
percent 100;
buffer-partition lossless {
percent 5;
}
buffer-partition lossy {
percent 90;
}
buffer-partition multicast {
percent 5;
}
}

Is there any experience or any best practice to do this? Thanks for your
help!


-- 
BR!



   James Chen
___
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Re: [j-nsp] Reset DSCP on ingress (again)

[Victor Sudakov]

Dale Shaw wrote:
> >
> > What is the easiest method of resetting the dscp value in incoming
> > packets when the incoming port is configured for "family
> > ethernet-switching"?
> >
> > If the port were in "family inet", I would use the firewall, but the
> > "family ethernet-switching" firewall does not seem to have a means to
> > set dscp.
> >
> > Thanks a lot in advance.
> >
> > The device is EX4200-24T, 12.3R6.6
> 
> Ingress DSCP marking isn't supported on EX4200.

Thanks for pointing this out.

> 
> You need to classify (into a FC) on ingress, and rewrite on egress. You can
> use a simple fixed classifier on ingress but packet marking/rewrite is only
> performed on egress.
> 
> Example of a fixed classifier:
> 
> # set class-of-service interfaces ge-0/1/0 unit 0 forwarding-class
> best-effort

Is a fixed classifier supported on a vlan interface?

-- 
Victor Sudakov,  VAS4-RIPE, VAS47-RIPN
AS43859
___
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp