FIFO FQ WFQ CBFWQ

FIFO FQ WFQ CBFWQ

Overview of Ques

· two main sections you have the software que and the hardware que (tx ring)

· THe hardware que or the tx ring is where the packets are put just before they are serialised

· We have no QOS control over the hardware que. It is very smallque.

· If this que is full the packets are buffered and qued in the software que.

· The hardware que is alway FIFO

· If hardware que is full traffic is qued in software

· With QOS we can control the configuration and processing of the software ques including  which traffic recieves priority

Queing Mechanism

FIFO

· The simplest software queing mechanism is FIFO or first in first out.

· packets are qued in  the order they are recieved and proces like this.

· This is the default on above 2mbs links.

Commands

no fair-que - turning off fair queing will enable fifo

sh int queing- view queing method

hold-que out - configure size of que

FAIR QUE (FQ)

· Also known as max-min sheduling

· Dynamically creates a que for each flow.

· Each que is given equal precendence as regard bandwidth allocation

· Each que is fifo and has CDT discard threshold which is the number of packets in the que before it will start tail dropping them

· The idea is that if i have low bandwidth application my que will never reach it CDT where if i have bandwidth hungry application trying to eat up the bandwidth that que will drop traffic. So the lower sender is never starved out by the high bandwidth sender

Commands

fair-que

Weighted Fair Queing

· is basically fair que with weighted allocation.

· by default weight is based on precedence DSCP so if my flow packets have a higher precendence than another flow we will get more weighting as regard bandwidth allocation

· It also use per flow fifo

· WFQ has a fixed number of ques the que number is based on bandwidth e.g 64 kbps has 16

· if you go over your que allocation flows are merged into same que so obviously not as effective if you go over

Commands

 

fair-que  and optional CDT QUEUES RSVP

policy-map WFQ

class class-default

fair-queue

int fa0/0

service-policy WFQ

In the newer IOS

show queue s0/1/0

has been replaced with

sh policy-map interface

Class Based Weighted Fair Queing/HQF

· This is weighted fair queing with manual defined classes so the network admin manually defines the class and can decide the allocation that is given to each class

· So you can manual guarantee bandwidth for certain traffic types

· By default it will tail drop if over conjested 

· Class default is always guaranteed 1% as minimum of the bandwidth

· max reserved bandwidth use to be 75 % but this has been depreciated in newer ios

· When you guarantee bandwidth in this manner precedence and dscp will not be used for quieing

Commands

We can Reserve ir assign the bandwidth in a few ways

Abosulte bandwidth "kbps"

or

Relative bandwidth percent %  - the obvious advantage of relative it can be applied to any

class-map telnet

match protocol telnet

class-map http

match protocol http

class-map smtp

match protocol smtp

policy CBWFQ

class telnet

bandwidth percent 10

class http

bandwidth percent 20

fair-que

class smtp

bandwidth percent 20

int s0/0

service-policy out CBWFQ

so in this example telnet gets at a minimum 10 % of the bandwidth and will use fifo queing

while http is getting at minmum 20 % and within this que is fair queing

The default class will get the remainder so 50 % in this case. If http is not using it 20 %

telnet may make use of it is not completely reserved just sitting there waiting it is more of

weighting on the traffic than anything else