Basic Router Forwarding Question

Unanswered Question
Jun 28th, 2010
User Badges:

Hello, people..

Here is a silly question that I am amazed I dont have an answer for.

If a router has a 1Gbps LAN connection and all the outbound LAN traffic is heading out the router's serial interface, which is only a T1, how does the router handle the disparity in speeds?

In other words, lets say the router is receiving LAN traffic at a rate of 800 Mbps, and all of it is going out the serial interface, which is only 1.5 Mbps, the router will have to buffer a lot of traffic....right? It may drop a lot of it, correct? Lets imagine no QoS applied, just a vanilla set up.


  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 4 (1 ratings)
manish arora Mon, 06/28/2010 - 12:15
User Badges:
  • Silver, 250 points or more

If i remember correctly , it will store the packets in the hardware as well as software queues till they fill up and then will start tail dropping. The best was to avoid this to place shaping or policing on the Lan interface , ILF on the external interface. there is a lot that can be done to avoid packet loss and maintain some kind of packet qos. but yes , when the queues get filled up then it will cause tail dropping.

ex-engineer Mon, 06/28/2010 - 16:18
User Badges:

So I guess the router's interface buffers will store the data before it can be processed by the routers switching and forwarding engine.

Any other input?


Hitesh Vinzoda Mon, 06/28/2010 - 20:57
User Badges:
  • Silver, 250 points or more


There is basic explanation as below :

Data >> Software queue's (Buffer) >> Hardware queue or Tx ring.

When the data is received in high bandwidth interface and intended go out of slow wan link they are queued in software queue's till the queue is full, then you will experience tail drop in the queue, the packet from queue are then send to Tx ring. Tx ring is the last point from where packet is transmitted on wire. you may tweak the software queueing strategies to meet you requirement. few types of queueing are





LLQ etc.

Also Shaping strategy can be implemented to meet the output rate of the interface.


Hitesh Vinzoda

Pls rate useful posts.

Tharak Abraham Tue, 06/29/2010 - 02:28
User Badges:
  • Bronze, 100 points or more

Its not a silly question,

Its how it happens in real world.

You eat one spoon at a time out of ten spoonful food, everything is fine.

If its ten spoonfuls at a time, you wont be able to chew properly, food will drop out of your mouth and the entire process takes more time...-:)

Policing and Shaping are the tools for this purpose.

Policing rate limits the flow to a specified rate either inbound or outbound.

Shaping does the job in the outbound direction.

Shaper would have the so called fancy queueing mechanisms to give importance to flows but again limited based on the size of the queue.

Anything more than the queue can handle will be dropped (tail drop)


This Discussion