Traffic Shaping on Cisco IOS

Hi Rene,

Thanks for the QoS explanation. How would I implement in a broadband (PPPoE) environment where radius assigns the up/down bit rate for users?

Rohan

Hi Rohan,

You mean on the client side? Or on the provider side where you can attach a policy for each PPPoE session?

Rene

Hello Rene,

Yes. On the provider side.

Rohan

Rene,

Apparently the attributes returned by Radius will be assigned automatically to PPPoE sessions once the policy is not configured directly on any interfaces.

R1(config)#policy-map SHAPE_AVERAGE
R1(config-pmap)#class IPERF
R1(config-pmap-c)#shape average 64000

Radius attribute:
sub-qos-policy-in = SHAPE_AVERAGE

Rohan

1 Like

@ReneMolenaar @lagapidis @andrew Hi. I need to simulate WAN environment in lab. Precisely: need to increase delay between to max (200 ms let’s say) from what I have now 1/2 ms!!! Is this possible anyway between Cisco routers without using external simulator?

Hello Deep

There are a few of ways that come to mind that you could do this without the use of a simulator. These are not quite as elegant as would be desired, but it’s worth a shot.

The first is to increase the processing time for the packets. You could for example use a second server to generate extra traffic and then shape the interface to a low value. This would delay the packets but not drop them. You are limited by the interface queue sizes so you may end up dropping traffic before you introduce enough delay to the target flow, but it’s worth a shot.

Secondly, if you have a couple of spare routers with serial ports, you can connect them together and place them between your source and destination. You can configure a low clock rate and introduce a bottleneck.

Thirdly, you could use a PC with two network cards and use an open source firewall software such as Monowall where you can configure a pipe and set the latency you desire for traffic.

I hope this has been helpful!

Laz

1 Like

Thanks Laz for the reply. I could easily create more than 200ms delay in Linux but was not that easy in cisco. I was getting loss when tuned the interfaces aggressively earlier. I limited the BW on on interface (policed) and applied outbound shaping (its only outbound anyway) to get more than 100ms delay on directly connected interfaces. It works. Thanks a lot.

1 Like

Hello Rene/Laz,
I have a question and I am going to use the below topology for my question.
image

We have a few remote sites that are connected to the headquarter through private circuits. Now the thing is that we are increasing our bandwidth on each remote site. Therefore, I need to increase the bandwidth in the QoS Shaper on edge routers on each site and test the bandwidth to make sure the service provider has increased the bandwidth on their side. I was thinking to install Jperf on a remote site PC to test the bandwidth and it seems like that is not a very convenient solution. I have to come up with something that is easier and faster since I have to do whole bunch of sites. I also thought about running speed tests on remote site PCs, but speed test is not accurate all the time. Is there any test that I can run from the remote site routers to test the bandwidth?

Thanks in advance.

Azm

Hi Azm,

On a router, you could use IP SLA or TTCP but they don’t really offer any real throughput. You can’t really use those tools to test your bandwidth.

I think your best option is iperf, not jperf…no need to hassle with a GUI and java. With iperf, you can set up a receiver on your HQ and then paste the same command on a PC at each branch router to transmit.

It’s not as ideal as using a router as the transmitter but it does allow you to actually test the bandwidth up to a high rate.

Rene

1 Like

Hello Rene,

I have a 3-tier MQC config example here, but I don’t know how to appropriately interpret it and have some questions to ask.

In the below example, in level-1 police-map:
- I understand it’s shapping the rate to 10Mbps with a Bc of 40000bps and Be of 0. But what is the command “account” doing?

In the below example, in level-2 police-map:
- The command priority is for LLQ, but if no ammount of bandwidth is specified, how does it get priority?
- What’s the reason to create a queue-limit? Any negative impact if don’t create a queue-limit?
- How does this command “bandwidth remaining percent 60” mean? Can we use upto 60% or 40%?
- Is this the bandwidth % under the level-1 policy-map’s shapiing rate ? or the actual speed of the interface?
- Can you explain briefly the functions of those 4 commands of “random-detect” ?

In the below example, in level-3 police-map:
- What’s the default action if no violated-action is defined ?

====Config example====

interface GigabitEthernet0/0.5
 bandwidth 9856
 encapsulation dot1Q 5
 ip address 10.1.1.1 255.255.255.252
 no cdp enable
 service-policy output Level1

policy-map Level1
 class class-default
  shape average 10000000 40000 0 account user-defined 28
   service-policy Level2

policy-map Level2
  class Level3-Voice-Traffic
  priority
  queue-limit 2048 packets
   service-policy Level3-Voice
 class Level3-any-traffic
  bandwidth remaining percent 60 account user-defined 28
  queue-limit 1024 packets
  random-detect dscp-based
  random-detect exponential-weighting-constant 1
  random-detect dscp 26 384 512 10
  random-detect dscp 28 256 384 10
   service-policy Level3-any

policy-map Level3-Voice
 class Level3-Voice-Traffic
  police 4000000 500000 500000 conform-action set-dscp-transmit ef exceed-action drop
policy-map Level3-any
  class Level3-Any-Traffic
   police 6000000 750000 750000 conform-action set-dscp-transmit af31 exceed-action set-dscp-transmit af32

class-map Level3-Voice-Traffic
 match access-group name Voice-Traffic
class-map Level3-Any-Traffic
 match any

ip access-list extended Voice-Traffic
 permit ip any any dscp ef
 permit ip any 10.150.0.0 0.0.255.255

==============================

Regards,
Ray

Hello Laz,
I am trying to configure a port on a 3850 as the routed port that is going to an upstream service provider private cloud. If I configure QoS shaper on the routed uplink port, how is it going to behave?

I have another question and I am going to use the below configuration as the reference. What would be the impact of using fair-queue here? Someone said, if I use fair-queue, I am not going to get the full 100 Mb bandwidth because fair-queue will cause flow drop. Would you please explain it to me?

policy-map WAN_LINK
 class VOICE
  priority percent 33

 class NETWORK
  bandwidth percent 7
 
 class CRITICAL-APP
  bandwidth percent 35 
  fair-queue
  random-detect dscp-based

 class class-default
  bandwidth percent 25 
  fair-queue
  random-detect dscp-based

policy-map 100Mb_SHAPER
 class class-default
  shape average 100000000
   service-policy WAN_LINK

Thanks in advance.

Best regards,
Azm Uddin

Hello AZM

An excellent lesson that will probably have all the info you need for this specific implementation is the following:

Look it over and if you have any more specific questions, please feel free to ask!!

As for the second question, the fair-queue command specifies the number of dynamic queues to be reserved for use by the class-default class as part of the default class policy. The number of queues can be specified, but if it is not as in your example, a default value is defined depending on the bandwidth of the interface.

For more information about this command and its parameters, take a look at this Cisco documentation.

When used along with the bandwidth command, as it is in the example you gave, then Cisco gives the following explanation:

If a default class is configured with the bandwidth policy-map class configuration command, all unclassified traffic is put into a single FIFO queue and given treatment according to the configured bandwidth. If a default class is configured with the fair-queue command, all unclassified traffic is flow classified and given best-effort treatment. If no default class is configured, then by default the traffic that does not match any of the configured classes is flow classified and given best-effort treatment. Once a packet is classified, all of the standard mechanisms that can be used to differentiate service among the classes apply.

(Taken from here)

I hope this gives you some guidance as to the use of the command and the affect it will have in the specific configuration.

I hope this has been helpful!

Laz

1 Like

Hello Laz,
I am sorry for the confusion. I was asking the question from shaping performance perspective. Since I am configuring shaper on a 3850 routed port, is the 3850 routed port going to perform shaping and queuing the same way as a router like ISR 4331 or 4321 would do or since it is a switch, it does not have the full blown shaping and queuing capability like a router?

Best Regards,
Azm Uddin

Hello AZM

I don’t have any direct experience with the differences in performance between queuing and shaping mechanisms on a routed port on an L3 switch and similar configurations on an ISR. The methodology and logic behind the configuration in each case is similar, but the performance will be more affected by the resources provided by each platform such as memory available for each queue, ASIC architecture and raw CPU power. Some of the new 6800 series Catalyst switches have exceptional hardware for queuing providing up to 500 MB (yes megabytes) buffers for queuing per port!

I’ll give it over to @ReneMolenaar to answer more specifically for you as he has probably had more experience than me in these matters.

I hope this has been helpful!

Laz

Thank you so much Laz…

1 Like

What is the name of your terminal app that you using it in your videos ?

thank you

Hi Mohanad,

Nowadays, I use SecureCRT for everything. It’s not free but it looks good and clean, you can run (python) scripts through it, it has a decent session manager, etc.

A free alternative is xshell.

Rene

Oh ok
I have a paid version in my work so how you can change the background to the guad ruled
Like this ??

Hello Mohanad

SecureCRT is indeed quite versatile and can do quite a lot. In order to configure the colour scheme and backgrounds, you can check out this link from VanDyke systems, the creator of the software.

https://www.vandyke.com/support/tips/colorconfig.html

I hope this has been helpful!

Laz

Hi Team,

While using traffic shaping, suppose a class-map with default class is reserved with only 10% of bandwidth of 1G physical interface and at the same time there is no congestion occuring on the same physical interface, does the router drops packets if it exceeds the reserved bandwidth i.e. 10% of physical interface?
And if the queue-limit is increased to a maximum limit, does the excess traffic classified as default traffic gets bufferred even if the there is no congestion on the same interface?

Note: Platform considered here is Cisco IOS-XE and IOS-XR