Traffic shaping and Layer 2 overhead

Answered Question
Jul 16th, 2009

I have been puzzled by this for sometime and have been trying to get to the bottom of it.

I have 2 routers connected via a 10Mb FD Ethernet link and on router 1 I am traffic shaping outbound on this link with the following config:

Policy Map SHAPE

Class class-default

Traffic Shaping

Average Rate Traffic Shaping

CIR 64000 (bps) Max. Buffers Limit 1000 (Packets)

Bc 1280 Be 0

When I flood this link I see the following on show interface f0/0:

poc#sh int f0/0 | inc 30 second

30 second input rate 0 bits/sec, 3 packets/sec

30 second output rate 64000 bits/sec, 144 packets/sec

However on the other routers interface I see:

spray#sh int f6 | inc 30 second

30 second input rate 75000 bits/sec, 148 packets/sec

30 second output rate 0 bits/sec, 0 packets/sec

So I am confused as to why interface commands show a difference in data i.e 64Kb on the outgoing interface (poc) and 75Kb on the incoming interface (spray)?

This is a back to back connection between 2 seperate routers. (poc out connected to spray in)

And what is this 9-10 bytes per packet of unaccounted traffic?

I have this problem too.
0 votes
Correct Answer by Giuseppe Larosa about 7 years 4 months ago

Hello Kate,

interface counters on ethernet interfaces consider the layer2 frame without the final 4 byte FCS so the size of frame is seen as

IP size + 14 bytes (but it cannot be less then 60 bytes because ethernet frame mimimum size is 64 bytes, 1518 bytes max including FCS)

I'm sure old CAR committed access rate works taking in account L2 overhead and I think also Class based shaping and policing do the same.

Hope to help

Giuseppe

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
Loading.
Correct Answer
Giuseppe Larosa Thu, 07/16/2009 - 09:29

Hello Kate,

interface counters on ethernet interfaces consider the layer2 frame without the final 4 byte FCS so the size of frame is seen as

IP size + 14 bytes (but it cannot be less then 60 bytes because ethernet frame mimimum size is 64 bytes, 1518 bytes max including FCS)

I'm sure old CAR committed access rate works taking in account L2 overhead and I think also Class based shaping and policing do the same.

Hope to help

Giuseppe

aniazik1981 Fri, 09/02/2011 - 09:43

Hi Giuseppe,

Recently I have came across similar question as Kate.

We have a WAN ASR1006 Cisco router connected over Ethernet WAN interface to the remote location.

interface GigabitEthernet1/2/3

bandwidth 200000

ip address x.x.x.x 255.255.255.252

ip flow ingress

ip flow egress

logging event link-status

load-interval 30

no negotiation auto

cdp enable

service-policy output XXXXXX

end

We have a shaping policy applied to the interface as follows:

GigabitEthernet1/2/3

Service-policy output: XXXXXXX

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

329736629 packets, 261478270158 bytes

30 second offered rate 94461000 bps, drop rate 24000 bps

Match: any

Queueing

queue limit 791 packets

(queue depth/total drops/no-buffer drops) 0/85300/0

(pkts output/bytes output) 329743966/261475845041

shape (average) cir 190000000, bc 760000, be 0

target shape rate 190000000

We have a Netscout TAP on the circuit which feeds data into Corvil monitoring device.

The Corvils has been configured for 4 ms sampling (same as the Tc configured on Cisco router). The max traffic volumes reported by the Corvil are higher than 190Mbps and on avarage max traffic volumes for 4 ms sampling are around 194Mbps. And this could be explained by what you said that the router doesn't count Ethernet 4bytes FCS while it calculates Bc (commited burst size) for shaping purposes.

However few times a day I can see a spike of a 220-230Mbps, which is imposible to have on 200Mbps.

Is there any explanation of that or this is more likely to be Corvil monitoring device inaccuracy/mistake.

I have noticed as well that the traffic which seems to cause this spike is UDP video traffic.

Thanks in advance.

Anna

Phil Leinster Wed, 11/28/2012 - 04:54

Hi Giuseppe,

What you say makes sense, however can you explain the discrepancy on the packets per second count? (144pps vs. 148pps)

Thanks,

Phil



Actions

This Discussion