QoS Traffic Shaping Explained

Hi Rene,

Excellent explanation.

By doing the BE, will customer feel any different since only first internal allow to burst extra bit?

Davis

Hi Davis,

If you have long transfers then you will hardly notice it. For shorter transfers you might notice a bit of an improvement.

Rene

Hi Rene,

Yea, i agree with you.

Thanks

Davis

Hi Rene,

Excellent article.

I like to come back to the statement, that packets, which does not fit into the bc or bc+be, are dropped.

  1. why is that? Is the endpoint not able to wait for the remaining bits with the next time intervall in order to have the complete packet?

  2. Assuming I have a packet, which is in general larger than bc or bc+be. According to the statement it will never make it?

So I have to use fragmentation?

Thanks for your help.

Hi Tamas,

When you use shaping, it will be buffered. The policer will drop packets if it is exceeded. When you configure a policer with a small BC/BE then yes, you’ll see that all packets that exceed the size of the BC/BE will be dropped.

They will be dropped since that’s what we have configured :slight_smile:

Rene

Hi Rene,
Can you please explain why shaping can be applied on an interface only on outbound vs. policing that can be also applied inbound and outbound? I have read several materials but it’s still very cloudy to me. Thank you!
A.

Hi Adrian,

Shaping is basically queueing. We put packets in the queue and only send traffic every now and then to get the desired shaping rate.

This isn’t really possible inbound since you already have received the traffic. I guess you could store it in an inbound queue but that might make not much sense to do :slight_smile:

Rene

Silly question. Is the Be usually the same as the Bc or can the Be value be different?

Michael,
Be can be the same as Bc, but is often different. For example, with a single rate, three color policer, Bc and Be are the same, but with a dual rate, three color policer, Bc is based on the Committed Information Rate (CIR), while Be is based on the Peak Information Rate (PIR) which are often different.

Hi Rene,

Talking about bitrate and “Routers are only able to send bits at the physical clock rate”, I was confusing about the very basic theory of the packet, packet size and how it transmit in the network media.

If everything is sending bit by bit, how it become a “packet” and people always says how the data in a message or file is broken up into packages about 1,500 bytes long, but how it change to “a packet”? so router is sending bit or sending packet?

Thanks

Hi,
If i have one gb bandwidth and the source transmitting only 5 Mb,still do I need to work on qos
thanks

Sims,
It may seem crazy in your scenario that you would have to worry about QoS, but yes, you do, if you have mission critical applications that can’t afford any type of delay. There is a sneaky phenomenon called a “micro-burst” where a sudden, very short (milliseconds) flood of traffic can overwhelm a device. If you have a mission critical application, it would be smart for you to establish a QoS policy that guarantees the bandwidth you know it needs as opposed to leaving it to chance.

19 posts were merged into an existing topic: QoS Traffic Shaping Explained

Please help me out to implement QoS over ADSL link. I want to prioritize my voice traffic and assign 15% of BW.
Please check the config of my usual WAN link not sure what would be on ADSL.

class-map match-any Client-Signal-Class
 match precedence 3
 match ip dscp af31
 match ip dscp cs3
class-map match-any Client-VOIP-Class
 match precedence 5
 match ip dscp ef
!
policy-map 50Mb-VOIP-PMAP
 class Client-VOIP-Class
  priority percent 10
 class Client-Signal-Class
  bandwidth percent 5
 class class-default
  fair-queue
policy-map 50Mb-VOIP
 class class-default
  shape average 49000000 504000
   service-policy 50Mb-VOIP-PMAP

Not sure what would be the shaping reference BW (Here - 49000000) on ADSL link. Your prompt reply highly appreciated.

This example looks correct. You are shaping to 49 Mbit, with a Bc of 504000 you have a Tc of ~11ms which is fine. On Cisco IOS, you have to configure the shaper first and then use a sub-policy with LLQ like you did…10% priority queue for RTP and 5% (CBWFQ) for your signaling traffic.

Hi,

Rene, can you tell me if the Be is available at any given time or not ? If i have Bc 8000bits and Be 8000 it means that at every Tc i can send 16 000bits ?

Hello Bogdan

The Be will be available only if it has been configured. If it has, then for the first Tc you will be able to send up to 16000 bits. In order for the Be to be available again, a certain time interval of inactivity, the Be will be available again to burst up to 16000 bits once again.

I hope this has been helpful!

Laz

Hello Rene

I am bit confuse between policing and shaping … can you please help to understand practical difference between policing and shaping …

Hello Prashant,

Both policing and shaping have a common goal: to “rate-limit” exceeding traffic.

How they do it is different though. A shaper will “buffer” the traffic while a policer “drops” the traffic.

A practical example could be an ISP router that is connected to multiple customer routers. Let’s say that the routers are connected with GigabitEthernet links but the customers are only paying for a 100/100 Mbit connection.

The ISP will then use a policer to drop all incoming traffic that exceeds 100 Mbit.

On the customer end, you probably don’t want your traffic to get dropped so you can then configure an outgoing shaper that is configured for 100 Mbit. This will not drop your traffic but buffer it, which guarantees that all traffic is sent and the ISP’s policer won’t drop it.

Hope this helps!

Rene

1 Like

Is there a way to see how much buffer space is being used, either globally or per interface, and any way to adjust this?