Re: Need help to understand QoS configuration

CSCO11544512 · ‎10-28-2013

Hi,

I need help to understand this configuration. Could you help me ?

class-map match-any DSCP-IN-D1INPUT

match access-group name DSCP-IN-D1INPUT

class-map match-any DSCP-IN-D2INPUT

match access-group name DSCP-IN-D2INPUT

class-map match-any DSCP-IN-D3INPUT

match access-group name DSCP-IN-D3INPUT

class-map match-any DSCP-IN-D4INPUT

match access-group name DSCP-IN-D4INPUT

class-map match-any DSCP-OUT-D4-OFFICE

match ip dscp af21

class-map match-any DSCP-OUT-D1-OFFICE

match ip dscp ef

class-map match-any DSCP-OUT-D2-OFFICE

match ip dscp af41

class-map match-any DSCP-OUT-D3-OFFICE

match ip dscp af31

class-map match-all DSCP-COUNT-D4INPUT

match ip dscp af21

class-map match-all DSCP-COUNT-D1INPUT

match ip dscp ef

class-map match-any DSCP-COUNT-D2INPUT

match ip dscp af41

class-map match-all DSCP-COUNT-D3INPUT

match ip dscp af31

class-map match-all DSCP-OUT-D1INPUT

match ip dscp ef

class-map match-all DSCP-OUT-D3INPUT

match ip dscp af31

class-map match-all DSCP-OUT-D2INPUT

match ip dscp af41

class-map match-all DSCP-OUT-D4INPUT

match ip dscp af21

!

policy-map COS-OUT-D4-OFFICE

class DSCP-OUT-D4INPUT

police cir 300000 bc 150000 be 75000

conform-action transmit

exceed-action set-dscp-transmit af12

policy-map COS-OUT-D3-OFFICE

class DSCP-OUT-D3INPUT

police cir 2400000 bc 600000 be 300000

conform-action transmit

exceed-action set-dscp-transmit af22

policy-map COS-OUT-D2-OFFICE

class DSCP-OUT-D2INPUT

police cir 2700000 bc 600000 be 300000

conform-action transmit

exceed-action set-dscp-transmit af32

policy-map COS-OUT-D1-OFFICE

class DSCP-OUT-D1INPUT

police cir 600000 bc 37000 be 14000

conform-action transmit

exceed-action set-dscp-transmit af42

policy-map COS-OUT-OFFICE

class DSCP-OUT-D1-OFFICE

bandwidth remaining percent 5

random-detect dscp-based

random-detect dscp 26 25 75 20

random-detect dscp 28 12 24 20

service-policy COS-OUT-D1-OFFICE

class DSCP-OUT-D2-OFFICE

bandwidth remaining percent 40

random-detect dscp-based

random-detect dscp 18 38 76 20

random-detect dscp 20 21 42 20

service-policy COS-OUT-D2-OFFICE

class DSCP-OUT-D3-OFFICE

bandwidth remaining percent 45

random-detect dscp-based

random-detect dscp 10 48 96 20

random-detect dscp 12 31 62 20

service-policy COS-OUT-D3-OFFICE

class DSCP-OUT-D4-OFFICE

bandwidth remaining percent 9

random-detect dscp-based

random-detect dscp 10 48 96 20

random-detect dscp 12 31 62 20

service-policy COS-OUT-D4-OFFICE

policy-map COUNT-IN

class DSCP-COUNT-D1INPUT

set dscp ef

class DSCP-COUNT-D2INPUT

set dscp af41

class DSCP-COUNT-D3INPUT

set dscp af31

class DSCP-COUNT-D4INPUT

set dscp af21

policy-map COS-IN

class DSCP-IN-D1INPUT

set dscp ef

class DSCP-IN-D2INPUT

set dscp af41

class DSCP-IN-D3INPUT

set dscp af31

class DSCP-IN-D4INPUT

set dscp af21

!

interface Multilink13

! Extract from the full configuration

service-policy INPUT COUNT-IN

service-policy output COS-OUT-OFFICE

!

interface FastEthernet0/1

! Extract from the full configuration

service-policy INPUTut COS-IN

Regards

Joseph W. Doherty · ‎10-28-2013

What kind of help do you need? Reason I ask, learning QoS can be a multi-day course, so was this something in particular you don't understand?

Also, what's the device and its running IOS version?

Vasilii Mikhailovskii · ‎10-28-2013

Hello.

The logic here is:

f0/1 - for inbound traffic service-policy is doing marking... ACL DSCP-IN-D1INPUT marked as EF, DSCP-IN-D2INPUT marked as af41 and etc.
input service-policy on M13 is doing nothing, but counting incoming traffic (no idea what is it for );
output service-policy on M13 is not obvious...

The primary aim of that policy is to limit class bandwidth to some predefined values... at the same time doing remarking in case what bandwidth exceeded, but bandwidth assigned to class is no consistent with policer values

So:

EF class is targeted to 600K, remarking to af42;
AF4 - 2700K, remarking to af32;
AF3 - 2400K, remarking to af22;
AF2 - 300K , remarking to af12.

Configuration is not consistent sometimes... for example, EF remarked into af42, but queue size is tuned for af31/af32.

I would say it's easier to rewrite the policy, like following:

policy-map COS-OUT-OFFICE

class DSCP-OUT-D1-OFFICE

bandwidth remaining percent 5

police cir 600000 bc 37000 be 14000

conform-action transmit

exceed-action set-dscp-transmit af42

random-detect dscp-based

random-detect dscp 46 25 75 20

random-detect dscp 36 12 24 20

class DSCP-OUT-D2-OFFICE

bandwidth remaining percent 40

police cir 2700000 bc 600000 be 300000

conform-action transmit

exceed-action set-dscp-transmit af32

random-detect dscp-based

random-detect dscp 34 38 76 20

random-detect dscp 28 21 42 20

class DSCP-OUT-D3-OFFICE

bandwidth remaining percent 45

police cir 2400000 bc 600000 be 300000

conform-action transmit

exceed-action set-dscp-transmit af22

random-detect dscp-based

random-detect dscp 26 48 96 20

random-detect dscp 20 31 62 20

class DSCP-OUT-D4-OFFICE

bandwidth remaining percent 9

police cir 300000 bc 150000 be 75000

conform-action transmit

exceed-action set-dscp-transmit af12

random-detect dscp-based

random-detect dscp 18 48 96 20

random-detect dscp 12 31 62 20

class class-default

I would also guess that ACL DSCP-IN-D4INPUT defining scavenger class (due to low bandwidth)... but it's not obvous why do we lack bandwidth for default class.

I would replace "bandwidth remaining percent xx" with absolute values, as policer is configured with absolute rate.

If you need any more clarification, please share description of your and provider's QoS policy, as well as content of ACLs in use.

Please sare output of show policy-map out M13

CSCO11544512 · ‎10-29-2013

Hi,

I was looking for this kind of explanation !

The IOS : c2800nm-ipbasek9-mz.124-24.T3.bin

Here is a short explanation by the author of the QoS (Google Translate)

- When a stream enters the network, whether from a server to a PC or phone, it passes through an access-list that identifies it.

- It is then marked by the router as: real-time priority, normal, best effort by example

- All devices receive the package later can read this value (dscp)

- Each router will then be on its WAN interfaces information about each queues and how to treat them.

- In case of congestion links, flows such as voice over IP, but also others, will therefore be transmitted first with guaranteed bandwidth on the WAN. flows lower priority will be delayed, see discarded in case of congestion.

The ACL :

ip access-list extended DSCP-IN-D1INPUT

permit ip 172.27.0.0 0.0.255.255 172.27.0.0 0.0.255.255

permit ip any host 172.31.2.92

permit ip any host 172.31.2.93

permit ip host 172.31.2.92 any

permit ip host 172.31.2.93 any

permit icmp any any

ip access-list extended DSCP-IN-D2INPUT

permit tcp any any eq 1494

permit udp any any eq 1604

permit tcp any any eq 3389

permit tcp any any eq www

permit tcp any any eq 443

permit tcp any any eq 1433

permit tcp any any eq 389

permit udp any any eq 389

permit tcp any any eq 88

permit udp any any eq 88

permit udp any any eq domain

permit udp any any eq netbios-ns

permit udp any any eq ntp

permit tcp any eq 1494 any

permit udp any eq 1604 any

permit tcp any eq 3389 any

permit tcp any eq www any

permit tcp any eq 443 any

permit tcp any eq 1433 any

permit tcp any eq 389 any

permit udp any eq 389 any

permit tcp any eq 88 any

permit udp any eq 88 any

permit udp any eq domain any

permit ip any host 172.31.5.201

permit ip host 172.31.5.201 any

permit ip any host 172.31.13.201

permit ip host 172.31.13.201 any

permit ip any host 172.31.137.201

permit ip host 172.31.137.201 any

ip access-list extended DSCP-IN-D3INPUT

deny ip any 172.25.0.0 0.0.255.255

deny ip 172.25.0.0 0.0.255.255 any

permit tcp any any eq 8080

permit tcp any any eq telnet

permit tcp any any eq 22

permit tcp any any eq 139

permit tcp any any eq 135

permit tcp any any eq 445

permit tcp any any eq 1025

permit tcp any eq 8080 any

permit tcp any eq telnet any

permit tcp any eq 22 any

permit tcp any eq 139 any

permit tcp any eq 135 any

permit tcp any eq 445 any

permit tcp any eq 1025 any

ip access-list extended DSCP-IN-D4INPUT

permit ip any 172.25.0.0 0.0.255.255

permit ip 172.25.0.0 0.0.255.255 any

permit tcp any any eq 9100

permit tcp any any eq lpd

permit tcp any any eq 2967

permit tcp any eq 9100 any

permit tcp any eq lpd any

permit tcp any eq 2967 any

permit ip any any

Output from sh policy-map out M13

The configuration may have changed

Service-policy INPUT: COUNT-IN

Class-map: DSCP-COUNT-D1INPUT (match-all)

850 packets, 85660 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: ip dscp ef (46)

QoS Set

dscp ef

Packets marked 850

Class-map: DSCP-COUNT-D2INPUT (match-any)

3320 packets, 724161 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: ip dscp af41 (34)

3320 packets, 724161 bytes

5 minute rate 0 bps

QoS Set

dscp af41

Packets marked 3320

Class-map: DSCP-COUNT-D3INPUT (match-all)

20172 packets, 3345732 bytes

5 minute offered rate 2000 bps, drop rate 0 bps

Match: ip dscp af31 (26)

QoS Set

dscp af31

Packets marked 20172

Class-map: DSCP-COUNT-D4INPUT (match-all)

164527 packets, 16400678 bytes

5 minute offered rate 72000 bps, drop rate 0 bps

Match: ip dscp af21 (18)

QoS Set

dscp af21

Packets marked 164527

Class-map: class-default (match-any)

3899 packets, 3348857 bytes

5 minute offered rate 44000 bps, drop rate 0 bps

Match: any

Service-policy output: COS-OUT-OFFICE

Class-map: DSCP-OUT-D1-OFFICE (match-any)

838 packets, 81476 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: ip dscp ef (46)

838 packets, 81476 bytes

5 minute rate 0 bps

Queueing

queue limit 64 packets

(queue depth/total drops/no-buffer drops) 0/0/0

(pkts output/bytes output) 838/81476

bandwidth remaining 5% (307 kbps)

Exp-weight-constant: 9 (1/512)

Mean queue depth: 0 packets

dscp Transmitted Random drop Tail drop Minimum

Maximum Mark

pkts/bytes pkts/bytes pkts/bytes thresh

thresh prob

af31 0/0 0/0 0/0 25

75 1/20

af32 0/0 0/0 0/0 12

24 1/20

ef 185676/17097363 0/0 0/0 36

40 1/10

Service-policy : COS-OUT-D1-OFFICE

Class-map: DSCP-OUT-D1INPUT (match-all)

838 packets, 81476 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: ip dscp ef (46)

police:

cir 600000 bps, bc 37000 bytes, be 14000 bytes

conformed 838 packets, 81476 bytes; actions:

transmit

exceeded 0 packets, 0 bytes; actions:

set-dscp-transmit af42

violated 0 packets, 0 bytes; actions:

set-dscp-transmit af42

conformed 0 bps, exceed 0 bps, violate 0 bps

Class-map: class-default (match-any)

0 packets, 0 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: any

Class-map: DSCP-OUT-D2-OFFICE (match-any)

3643 packets, 2688709 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: ip dscp af41 (34)

3643 packets, 2688709 bytes

5 minute rate 0 bps

Queueing

queue limit 64 packets

(queue depth/total drops/no-buffer drops) 0/0/0

(pkts output/bytes output) 3643/2688709

bandwidth remaining 40% (2457 kbps)

Exp-weight-constant: 9 (1/512)

Mean queue depth: 1 packets

dscp Transmitted Random drop Tail drop Minimum

Maximum Mark

pkts/bytes pkts/bytes pkts/bytes thresh

thresh prob

af21 0/0 0/0 0/0 38

76 1/20

af22 0/0 0/0 0/0 21

42 1/20

af32 10400/14631041 9/13168 50/72566 28

40 1/10

af41 624570/396758744 24/27796 166/213900 32

40 1/10

Service-policy : COS-OUT-D2-OFFICE

Class-map: DSCP-OUT-D2INPUT (match-all)

3643 packets, 2688709 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: ip dscp af41 (34)

police:

cir 2700000 bps, bc 600000 bytes, be 300000 bytes

conformed 3643 packets, 2688709 bytes; actions:

transmit

exceeded 0 packets, 0 bytes; actions:

set-dscp-transmit af32

violated 0 packets, 0 bytes; actions:

set-dscp-transmit af32

conformed 0 bps, exceed 0 bps, violate 0 bps

Class-map: class-default (match-any)

0 packets, 0 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: any

Class-map: DSCP-OUT-D3-OFFICE (match-any)

24989 packets, 21918576 bytes

5 minute offered rate 2000 bps, drop rate 0 bps

Match: ip dscp af31 (26)

24989 packets, 21918576 bytes

5 minute rate 2000 bps

Queueing

queue limit 64 packets

(queue depth/total drops/no-buffer drops) 0/0/0

(pkts output/bytes output) 24989/21918576

bandwidth remaining 45% (2764 kbps)

Exp-weight-constant: 9 (1/512)

Mean queue depth: 0 packets

dscp Transmitted Random drop Tail drop Minimum

Maximum Mark

pkts/bytes pkts/bytes pkts/bytes thresh

thresh prob

af11 0/0 0/0 0/0 48

96 1/20

af12 0/0 0/0 0/0 31

62 1/20

af22 146397/218231936 33/46270 20/29823 28

40 1/10

af31 3512639/2214458323 38/46246 49/57435 32

40 1/10

Service-policy : COS-OUT-D3-OFFICE

Class-map: DSCP-OUT-D3INPUT (match-all)

24989 packets, 21918576 bytes

5 minute offered rate 2000 bps, drop rate 0 bps

Match: ip dscp af31 (26)

police:

cir 2400000 bps, bc 600000 bytes, be 300000 bytes

conformed 24989 packets, 21918576 bytes; actions:

transmit

exceeded 0 packets, 0 bytes; actions:

set-dscp-transmit af22

violated 0 packets, 0 bytes; actions:

set-dscp-transmit af22

conformed 2000 bps, exceed 0 bps, violate 0 bps

Class-map: class-default (match-any)

0 packets, 0 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: any

Class-map: DSCP-OUT-D4-OFFICE (match-any)

165452 packets, 15493199 bytes

5 minute offered rate 89000 bps, drop rate 0 bps

Match: ip dscp af21 (18)

165452 packets, 15493199 bytes

5 minute rate 89000 bps

Queueing

queue limit 64 packets

(queue depth/total drops/no-buffer drops) 0/0/0

(pkts output/bytes output) 165451/15493114

bandwidth remaining 9% (552 kbps)

Exp-weight-constant: 9 (1/512)

Mean queue depth: 0 packets

dscp Transmitted Random drop Tail drop Minimum

Maximum Mark

pkts/bytes pkts/bytes pkts/bytes thresh

thresh prob

af11 0/0 0/0 0/0 48

96 1/20

af12 1011572/1505243377 102/153204 23/33098 31

62 1/20

af21 36146830/3404118332 10/7207 136/133603 32

40 1/10

Service-policy : COS-OUT-D4-OFFICE

Class-map: DSCP-OUT-D4INPUT (match-all)

165452 packets, 15493199 bytes

5 minute offered rate 89000 bps, drop rate 0 bps

Match: ip dscp af21 (18)

police:

cir 300000 bps, bc 150000 bytes, be 75000 bytes

conformed 164905 packets, 14673422 bytes; actions:

transmit

exceeded 50 packets, 74920 bytes; actions:

set-dscp-transmit af12

violated 496 packets, 744772 bytes; actions:

set-dscp-transmit af12

conformed 138000 bps, exceed 18000 bps, violate 173000 bps

Class-map: class-default (match-any)

0 packets, 0 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: any

Class-map: class-default (match-any)

748 packets, 68440 bytes

5 minute offered rate 0 bps, drop rate 0 bps

Match: any

queue limit 64 packets

(queue depth/total drops/no-buffer drops) 0/0/0

(pkts output/bytes output) 748/66564

Vasilii Mikhailovskii · ‎10-29-2013

Please provide your ISP's QoS settings/profile for the interface.

And what is the interface bandwidth?

Please paste output of show int mu13

PS: what other clarifications do you need?

CSCO11544512 · ‎10-29-2013

Hi

In fact, I don't understand how to choose these values (in bold) :

class DSCP-OUT-D1-OFFICE
    bandwidth remaining percent 5
    police cir 600000 bc 37000 be 14000
     conform-action transmit
     exceed-action set-dscp-transmit af42
     random-detect dscp-based
     random-detect dscp 46 25 75 20
     random-detect dscp 36 12 24 20

Here is the output from sh int Mu13

Multilink13 is up, line protocol is up

Hardware is multilink group interface

Internet address is 192.168.13.33/30

MTU 1500 bytes, BW 6144 Kbit/sec, DLY 100000 usec,

reliability 255/255, txload 2/255, rxload 2/255

Encapsulation PPP, LCP Open, multilink Open

Listen: CDPCP

Open: IPCP, loopback not set

Keepalive set (2 sec)

DTR is pulsed for 2 seconds on reset

Last input 00:00:01, output never, output hang never

Last clearing of "show interface" counters 10:46:41

Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 3

Queueing strategy: Class-based queueing

Output queue: 0/1000/0 (size/max total/drops)

5 minute input rate 58000 bits/sec, 60 packets/sec

5 minute output rate 57000 bits/sec, 59 packets/sec

3930200 packets input, 539608101 bytes, 0 no buffer

Received 0 broadcasts, 0 runts, 0 giants, 0 throttles

0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored, 0 abort

4026963 packets output, 942427089 bytes, 0 underruns

0 output errors, 0 collisions, 0 interface resets

0 unknown protocol drops

0 output buffer failures, 0 output buffers swapped out

0 carrier transitions

Vasilii Mikhailovskii · ‎10-29-2013

Hello.

class DSCP-OUT-D1-OFFICE

bandwidth remaining percent P0

police cir X1 bc X2 be X3

conform-action transmit

exceed-action set-dscp-transmit af42

random-detect dscp-based

random-detect dscp 46 Q1 Q2 D3

random-detect dscp 36 Q4 Q5 D6

X1 - target rate; X2 - is a bucket size... for real-time X2 should be about X1/100 or X1/50, for data classes it could be up to X1/20. X3 - it's to deal with bursts... it's not common to configure this value unless you own infrastructure, or pretty sure that ISP will proceed the exceeded traffic.

P0 should be a little bigger than X1 (otherwise policer won't help in case of congestion).

Q1 - it's a mean queue size when WRED will start dropping traffic with drop probability of 100% divided by D3.

Q2 - it's a mean queue size when all the new packet will be dropped ("tail drop").

Q4 much less than Q1 and Q5 than Q2 because we would like to drop exceeded (and remarked) traffic.

D3 and D6 - it's a drop probability (100% divided by D), but 20 is too much I guess; I would say D6 should be 5.

About D1 class: if it's IPT, then using WRED is not a good idea; IPT class should use LLQ (strict priority).

One more note about Q1/Q2.

Let's imaging all the queue is full with 1500 bytes packets... according to Q2, the queue size is 1500*96 = 144Kb; having only 300Kbps in case of congestion will result in "144000*8/300000= 3.84 sec of queuing delay... 48 is still too much, even if you count 800 byte packets!

That is cruel to printing class

PS: it's not common to have policer if your multilink is configured to survive in case of single link failture.

PS2: Q1/Q2/Q4/Q5/D6 (not sure if there is a reason to change D3) are chosen based on link bandwidth, link line speed (in your case it's the same), router's amount of memory, traffic pattern (requirement for latency, drops and etc.) and allocated (to class) bandwidth. In large implementations it's better to experiment with some sets of vlaues (and find the best based on end-user experience) before rolling out over the production.

Joseph W. Doherty · ‎10-30-2013

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Although I haven't carefully read all your replies, with just a quick glance at this last posting, my understanding of some of WRED's parameters is different.

The application to packet percentage, when queue's mean depth is at the first value is a little complicated to compute. You need to do a x/y line analysis, where the depth one less 1st-value is at zero percent, and percent is 1/3rd-value at the 2nd-value. At 2nd-value plus one, application percent is then 100.

It's also important to keep in mind, depth is a running average. It's possible actual instantaneous queue depth can be much larger, and there's no application of WRED percentages or even instantaneous queue depth is zero, and WRED impacts packets.

CSCO11544512 · ‎10-31-2013

Hi,

Thanks for all your answers.

The values for the random-detect dscp commands are a bit strange.

I still don't understand how don't understand why the DSCP-OUT-D1-OFFICE is designed for voice.

Regards

Joseph W. Doherty · ‎10-31-2013

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

I still don't understand how don't understand why the DSCP-OUT-D1-OFFICE is designed for voice.

IMO, it's not (well) designed for VoIP bearer.

Vasilii Mikhailovskii · ‎10-31-2013

The values for the random-detect dscp commands are a bit strange.

I still don't understand how don't understand why the DSCP-OUT-D1-OFFICE is designed for voice.

Please let me know why do you think they are "a bit strange"?

The class consumes voice traffic, as it matches all the traffic with dscp equal to EF.

CSCO11544512 · ‎11-01-2013

I don't understand why the third parameter is not simply equal to 100

Here is what I have found on an old Cisco documentation :
random-detect dscp dscp_value minimum-threshold maximum-threshold mark-probability-denominator

The values I find when I calculate the maximum probability of discard are very low.

And one naive question : if the routers are not configured for QoS, does the value of DSCP field have a great impact on traffic and drop ?

Joseph W. Doherty · ‎11-01-2013

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

I don't understand why the third parameter is not simply equal to 100

Because you're trying to avoid 100% drop. The whole point of RED is dropping some (i.e. as few as required) packets will cause the flow to slow its transmission rate. RED is trying to maximize "goodput".

And one naive question : if the routers are not configured for QoS, does the value of DSCP field have a great impact on traffic and drop ?

If the router isn't "looking" at ToS, DSCP has NO impact. (By default, most routers don't "look" at ToS.)