QOS, 3Mbit Internet on FastEthernet port?

Unanswered Question
Oct 10th, 2007

Hi all,

I have a 2811 router connected via one of it's FastEthernet ports to a managed Cisco router which provides Internet access for us.

I would like to use QOS to give traffic to certain key websites priority over other traffic. I am familiar with QOS over serial connections, and those work great. But yesterday a group of our students all launched BitTorrent and World of Warcraft, and completely saturated our 3Mb/sec of Internet bandwidth the entire day. I would have expected our important traffic to be given more bandwidth and less likely to be dropped, but even at it's worst, the queueing did not seem to work.

Can anyone help?

policy-map pmap_Merit_Out

class class_toMerit

set dscp af21

bandwidth remaining percent 75

class class-default

set dscp default

bandwidth remaining percent 25

class-map match-any class_toMerit

match access-group name ACL_dest_merit

interface FastEthernet0/0

service-policy output pmap_Merit_Out

During the height of the bandwidth problems this is what a "show queue" displayed; which is not at all what I would expect from a totally saturated link:

#show queue fastethernet0/0

Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0

Queueing strategy: Class-based queueing

Output queue: 0/1000/64/0 (size/max total/threshold/drops)

Conversations 0/2/256 (active/max active/max total)

Reserved Conversations 2/2 (allocated/max allocated)

Available Bandwidth 2304 kilobits/sec

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
lgijssel Wed, 10/10/2007 - 07:18

The most overlooked issue with managing Internet bandwidth is that the amount of outgoing traffic has no direct relation to the amount of incoming traffic. To state it simply: One mouseclick (=packet) will suffice to initiate a multi-megabyte download that will completely saturate the downstream link with the Internet for a long time. This seems to be exactly what is happening here as well.

It is not useful to police the outgoing (=upstream) traffic while your problem is with the downstream. You could rate-limit incoming traffic but that will not help you much; the traffic has already traversed the WAN-link (i.e. used bandwidth) and will be retransmitted by the sender when there is no acknowledgement. TCTIP is great in that! ;-)

When you look carefully at where the problem originates, the only place where policing the traffic would be useful is at the ISP-end of your Internet connection. Most ISP's are reluctant of this.



emphillips00 Wed, 10/10/2007 - 07:23

Hi Leo,

Thank you so much for your extremely fast and detailed reply! That is exactly what I was missing. I was focused on the speed differential (100 Mb/sec FastEthernet vs only 3 Mbit/sec of Internet), and completely overlooked the fact you mentioned.

That makes a lot more sense... In any case that I have used QOS successfully, it has been on a point to point link that I had control over both sides. But that is quite different than this case.

Thank you again for the great lesson!


Paolo Bevilacqua Wed, 10/10/2007 - 07:23


the problem is that the FE doesn't see congestion so the service-policy is ineffective. You need a "nested class map" so that traffic is first shaped at the 3 mbs that you can send to the internet, under it you will have your qos config that I *think* should work.

policy-map shape

class class-default

shape average 3000000

service-policy pmap_Merit_Out

interface FastEthernet0/0

service-policy output shape

hope this helps, please rate post if it does!


This Discussion