[j-nsp] RES: ACX2200 - bandwidth control at subinterfaces

[Alexandre Guimaraes]

Chris,
	I saw your email after this, don´t worry.
	
	I understand the situation where a routing platform had to be used
(MX routers). But, according ACX description, "a capable ACX Series
Universal Access Routers for metro-ethernet services" and so on....

	Perhaps my understanding of telecommunication is worst than ever
where in a world, the bandwidth is growing and growing faster than ever and
ACX provide good bandwitdth for small sites with a small price(less power,
less space, less price than a MX5/10), but these small sites had 50/100
l2circuits for different customers. In a 1Gig port, I can allocate 10/50
circuits giving low cost per port. (Thinking that everyone, had to monetize
their operation for each byte passing through).

	50 x 5mbps l2circuts cost less than 250Mbps peak, now imagine if 5
customers, with no bandwidth limits can do if  they reach 200Mbps each...
ACX is not for that? Only MX are? And those QFX? None of them had a minimal
bandwidth control or schedules or something that control the bandwidth.
Maybe I´m choose wrong my NNI equipment for small sites.

	I don´t know... 

		can someone help?




Alexandre

-----Mensagem original-----
De: juniper-nsp [mailto:juniper-nsp-bounces at puck.nether.net] Em nome de
Chris Kawchuk
Enviada em: quinta-feira, 25 de agosto de 2016 23:23
Para: juniper-nsp at puck.nether.net
Assunto: Re: [j-nsp] ACX2200 - bandwidth control at subinterfaces

You mean scheduler maps/shaping on a subinterface?

Correct.

EX doesn't do per-unit schedulers. If they did, nobody would buy an MX for
HQoS. ;)

You can do hard policers though... which is nasty.

I think you can still shape per-queue (i.e. [edit class-of-service
schedulers] best-effort shaping-rate XX;); so, using some output firewall
filters, you can put different VLANs into different queues, and shape each
queue. 

However you give up some of the QoS functionality (only 8 HQ queues to play
with...)


___________________________

sample config I hacked together in 5 minutes on an ex2200c - passes commit.


ge-0/0/0 {
    vlan-tagging;
    unit 1 {
        description "Some cusotmer - use best-effort HW queue - shape to
100m";
        vlan-id 1;
        family inet {
            address 1.1.1.1/24;
        }
    }
    unit 2 {
        description "another cusotmer - use assured-forwarding HW queue -
shape to 100m";
        vlan-id 2;
        family inet {
            address 2.2.2.1/24;
        }
    }
}


class-of-service {                      
    interfaces {
        ge-0/0/0 {
            scheduler-map my-wacky-per-queue-shaper;
        }
    }
    scheduler-maps {
        my-wacky-per-queue-shaper {
            forwarding-class best-effort scheduler best-effort-scheduler;
            forwarding-class assured-forwarding scheduler assured-scheduler;
        }
    }
    schedulers {
        best-effort-scheduler {
            shaping-rate 100m;
            buffer-size percent 50;
            priority low;
        }
        assured-scheduler {
            shaping-rate 100m;
            buffer-size percent 50;
            priority low;
        }
    }    

___________________________
                               

You'll need to use an output firewall filter on unit 1 to shove all traffic
into BE, and on unit 2 to shove all traffic to AF. Remember only 8 HQ
queues; and you'll likely reserve Queue 7 for network-control anyways.. so 7
effective queues (0-6) to play with.

Secondly, the EX has SMALL HW buffers; especially if I start carving them up
as I did above -- beware.

- CK.



On 25 Aug 2016, at 5:52 am, Alexandre Guimaraes
<alexandre.guimaraes at ascenty.com> wrote:

> Gents, afternoon,
> 
> 
> 	After some research and a talk with my SE about how to control 
> bandwidth at subinterfaces using ACX2200 Access Routers. I´h reached a 
> point where we can´t control bandwidth using subinterfaces.
> 
> 	Had someone of you guys, find a way to control that?
> 
> 	Class-of-services only control the interface itself, not the 
> subinterface.

_______________________________________________
juniper-nsp mailing list juniper-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp