cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1485
Views
10
Helpful
5
Replies

Understanding QoS mechanisms

danail-petrov
Level 1
Level 1

Hi guys,

i am preparing for Cisco 642-642 and now i read the "Cisco Press - IP Telephony Self-Study Cisco QOS Exam Certification Guide,Second Edition(2005)" book which is very good btw. But i'm little bit confused about how exactly QoS works in detail. Let's say that we have one router with two ethernet/fastethernet interfaces. We have one incoming and another outgoing interface:

(packet) ---> (router ingress (fa0/0) X router egress (fa0/1)) --->

In this case packet is arrived on fa0/0 interface, and here is my question. What is the order of the applied QoS mechanisms? I mean the packet has to be classified then it has to be policed, marked if we have CBQ (WFQ,CBWFQ) the packet needs to be ordered, then if we have shaper this packet have to be delayed (if it's necessarily of course). So i'm get confused of how and when packet getting queued (in order to achieve its place in fair queue), what is the sequence ? The packet first is queued and then is placed to shaping queue before it reach taildrop ? Can anyone explain me exactly (step-by-step) how the packet "rides" from end-to-end (ingress -> egress)? I'm confused of all these queues in different type of mechanisms. I mean, WFQ has it's queue, shaper manage it's own queue, we have hardware queue (TX/RX rings).. i believe that i'm getting understand all this, but i need just a little help maybe.

I'll be thankful to your answers!

Thanks in advance!

5 Replies 5

mohammedmahmoud
Level 11
Level 11

Hi,

"Cisco Press - IP Telephony Self-Study Cisco QOS Exam Certification Guide,Second Edition(2005)" is one of the best books indeed.

QoS tools are applied according to your configuration, as a best practice you shall do the classification and marking in the ingress as near to the source as possible, on the egress of the router you can configure the required QoS tool according to the class of the traffic and the issues that you are trying to mitigate, you should for example LLQ the voice traffic.

The only thing that you should care to is that how are the QoS tools applied, and i mean by that which tool can be applied inbound and which can be applied outbound, and the answer is that all tools are done outbound except Classification and Marking and Policing which can be done also inbound.

To summarize the QoS tools:

1. Classification and Marking - done at the input interface as near to the source as possible

2. Congestion Management (Queuing and Scheduling) - performed on the output interface.

3. Congestion Avoidance (Drop policies) - performed on the output interface.

4. Policing and Shaping - Policing can be done both inbound or outbound, but shaping can only be done outbound (since its a kind of queuing).

5. Link Efficiency Tools (Compression and LFI) - Done outbound for low speed links only (less than 768kbps).

> All techniques are done outbound except Classification and Marking and Policing which can be done also inbound.

HTH,

Mohammed Mahmoud.

Thank you Mohammed for your effort to answer my question. But what is really confusing me is the queues. If we use CBWFQ or just WFQ, we have queues right? So where is WRED applied? inside those queues(classes) or? I mean CBWFQ has 64 queues. The packet enter in queue in depend of ip precedence or dscp. So where is the WRED/RED mechanism is applied? Inside each one of these 64 queues? WRED has also weight right? By the way of logic WRED needs to be before WFQ to discard the scavenger traffic (with precedence/dscp/cos = 0 for example) and then the WFQ to schedule the packet? All this is confusing me. Can you please be more specific about exactly how this packet is treated? First is queued by the WFQ, then it's passed to the "big queue" where the RED/WRED choose it's destiny and then the packet is placed to TAIL-DROP or FIFO queue (which is actually TX buffer)?

Thank you once again and have a nice day!

Best Regards,

Danail Petrov

Hi Danail,

You are very welcomed, as i told you its all about how you are using the tools, for example in the case of CBWFQ you apply the WRED per class (and in CBWFQ, class = queue), another important thing, there is nothing called the big queue, the scheduler does the job to control the multiple queue per queuing tool, please do check this post:

http://forum.cisco.com/eforum/servlet/NetProf?page=netprof&forum=Network%20Infrastructure&topic=LAN%2C%20Switching%20and%20Routing&CommCmd=MB%3Fcmd%3Ddisplay_location%26location%3D.1ddfab9d

HTH,

Mohammed Mahmoud.

Thank you Mohammed!

I gave you 5 of 5. You really deserve it.

Hi Danail,

Thank you very much :), i am glade i've helped you out, and thank you very much for using the rating system.

BR,

Mohammed Mahmoud.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Innovations in Cisco Full Stack Observability - A new webinar from Cisco