Rate Limiting and QoS on 3750 WAN intf?

[apaxson]

All,

I'm currently using a 3750 in a co-lo environment as my co-lo backbone.

One of the interfaces goes to an MPLS cloud. Now, the co-lo delivers the circuit via Ethernet. As such, my interface is running 100Mbps. But, it's only a 3Mbps cloud.

How can I rate-limit this interface down to 3Mbps in such a way, to where I can apply my QoS policies, and QoS will use the rate-limited 3Mbps rather than the full 100Mbps.

According to the QoS Order of Operations, rate-limiting comes BEFORE queuing, and I'm afraid that my voice packets may get dropped.

I can police the individual voice traffic, but I want all other traffic to use up to 100% of the line, if voice is not there.

srr-queue bandwidth limit won't work by itself, as I can only go down to 10% (which, from 100Mbps is 10Mbps, and I need 3Mbps).

I guess I'm just getting hung up on the 3Mbps circuit. Since it's a 100Mbps interface, queuing will never really happen on my side, since the bandwidth will support it.

Thanks for any advice you have!

[Joseph W. Doherty · Accepted solution]

Can you run 10 Mbps with the MPLS provider's Ethernet?

If not, place a "dumb" L2 switch between your 3750 and MPLS Ethernet. Your side, 10 Mbps; MPLS side, 100 Mbps.

For either, once your interface is at 10, set ssr-queue bandwidth for 30%.

[apaxson]

josephdoherty,

As usual, I tend to make things more complicated than they need to be. :) I'm asking my provider for a 10Mb connection now.

Thanks for your assistance!

[kejeusa]

you could just manually set the bandwidth on the interface that would used by the qos policy.

int fa0/1

no switchport

ip address x.x.x.x

bandwidth 3000

[paolo bevilacqua]

The command "bandwidth" above is absolutely ineffective to set bandwidth as repeated countless times since the beginning of cisco.

[apaxson]

Correct. I believe the "bandwidth" command is only useful for routing protocols only, or if you are getting an interface speed from snmp to base graphs off of.

Thanks!

[Joseph W. Doherty]

Keenon, in this case, we're dealing with a 3750, so it doesn't work the same way it might on a router.

On routers (within platform/IOS limitations), we can use the bandwidth to control percentage based QoS.

e.g.

given this:

policy-map test

class class-default

shape average percent 50

On a Ethernet interface running at 100 Mbps, we get:

Rtr1#sh policy-map i o

FastEthernet0/0

Service-policy output: test

Class-map: class-default (match-any)

31 packets, 2076 bytes

5 minute offered rate 1000 bps, drop rate 0 bps

Match: any

Traffic Shaping

Target/Average Byte Sustain Excess Interval Increment

Rate Limit bits/int bits/int (ms) (bytes)

50 (%) 0 (ms) 0 (ms)

50000000/50000000 312500 1250000 1250000 25 156250

Adapt Queue Packets Bytes Packets Bytes Shaping

Active Depth Delayed Delayed Active

- 0 31 2076 0 0 no

But if we change the "bandwidth", e.g.:

Rtr1#sh run in fa 0/0

Building configuration...

Current configuration : 180 bytes

!

interface FastEthernet0/0

bandwidth 25000

We now get:

Rtr1#sh policy-map i o

FastEthernet0/0

Service-policy output: test

Class-map: class-default (match-any)

131 packets, 10654 bytes

5 minute offered rate 1000 bps, drop rate 0 bps

Match: any

Traffic Shaping

Target/Average Byte Sustain Excess Interval Increment

Rate Limit bits/int bits/int (ms) (bytes)

50 (%) 0 (ms) 0 (ms)

12500000/12500000 75000 300000 300000 24 37500

Adapt Queue Packets Bytes Packets Bytes Shaping

Active Depth Delayed Delayed Active

- 0 131 10654 0 0 no