QOS Precedence 4 bandwitdh

p.diaz · ‎08-27-2009

Hi.

I have the following problem.

A Client with a main and a remote site entering to an MPLS network. I have consistent QoS politics and we have voice traffic (precedence 5, priority queue) with not delay variation in the delay when congestion occurs. But when the video traffic circulates (prec 4, bandwitdh command) the delay in this traffic increases when cogestion occurs. Is this behavior normal? how can to avoid this behavior?

thanks.

Mohamed Sobair · ‎08-27-2009

Hi,

what kind of Queuing does class Video traffic has?

If its CBWFQ, then yes, as Bandwidth is guranteed, but delay is not Guranteed. If its delay sensitive, Configure LLQ for that class of traffic.

HTH

Mohamed

Joseph W. Doherty · ‎08-27-2009

Video is offen very variable in its bandwidth demand. It's possible your CBWFQ bandwidth statement provides enough bandwidth for your video's average bandwidth demand, but not enough to prevent queuing bursts. You could try increasing the bandwidth allocation. (Your video won't use any more average bandwidth, but a higher allocation means bursts are less likely to queue relative to other traffic.)

Another possible solution might be moving/converting the video to an LLQ class.

PS:

Since you mention MPLS, much can also depend on what your MPLS provider guarantees for IP Prec 4 traffic.

bmcginn · ‎08-27-2009

Hi there Pedro,

Can you post the QoS and interface configs?

Thanks,

Brad

p.diaz · ‎08-28-2009

(CE)-1-(PE)-2-(P)-3-(PE)-4-(CE)

(CE)1 (fast ethernet interface)

policy-map Shaping_Datos_FE

class class-default

shape average 3072000

service-policy QoS_Out

policy-map QoS_Out

class 6CoS_Prec5

priority 256

police cir 256000 bc 48000 be 96000

conform-action set-prec-transmit 5

exceed-action drop

class 6CoS_Prec4

set ip precedence 4

police cir 768000

conform-action transmit

exceed-action drop

class 6CoS_Prec3

bandwidth 512

set ip precedence 3

class 6CoS_Prec67

bandwidth 16

class 6CoS_Prec2

bandwidth 256

set ip precedence 2

class 6CoS_Prec1

bandwidth 256

set ip precedence 1

class class-default

set ip precedence 0

(PE)1 (Gigabit ethernet 100 Megas)

policy-map SHAPING3M

class class-default

shape average 3072000

service-policy CE

policy-map CE

class prec5

priority

police cir percent 9 (cir 276000 bps)

conform-action transmit

exceed-action drop

class prec4

bandwidth percent 25 (768 kbps)

class prec3

bandwidth percent 17 (522 kbps)

class prec2

bandwidth percent 9 (276 kbps)

class prec1

bandwidth percent 9 (276 kbps)

class prec67

bandwidth percent 1 (30 kbps)

2,3 10 GIGA LINKS WITH 10 PERCENT OF UTILIZACION AT MUCH

(PE)4

interface ATM0/0.5 point-to-point

description ROUTER 1841 BASICO - MAQUETA MPLS

bandwidth 1024

ip vrf forwarding vpn_12956_10000

ip address 10.19.10.22 255.255.255.252

no ip directed-broadcast

no atm enable-ilmi-trap

snmp trap link-status

pvc MQT 4/32

vbr-rt 1153 1153

encapsulation aal5snap

service-policy in QoS_In_ATM0/0.5

service-policy out QoS_Out_ATM0/0.5

policy-map QoS_Out_ATM0/0.5

class prec4

bandwidth percent 25

class prec5

police cir percent 10 conform-action transmit exceed-action drop

priority

class prec3

bandwidth percent 20

class prec2

bandwidth percent 15

class prec1

bandwidth percent 5

class prec67

bandwidth percent 2

(CE)4

map-class frame-relay FR_Out

frame-relay cir 1024000

frame-relay bc 10240

frame-relay mincir 1024000

service-policy input QoS_In

service-policy output QoS_Out

policy-map QoS_Out

class 6CoS_Prec5

priority 128

police cir 128000 bc 24000 be 48000

conform-action set-prec-transmit 5

exceed-action drop

class 6CoS_Prec4

set ip precedence 4

police cir 128000

conform-action transmit

exceed-action drop

class 6CoS_Prec3

bandwidth 128

set ip precedence 3

class 6CoS_Prec67

bandwidth 16

class 6CoS_Prec2

bandwidth 96

set ip precedence 2

class 6CoS_Prec1

bandwidth 128

set ip precedence 1

class class-default

set ip precedence 0

Joseph W. Doherty · ‎08-28-2009

Some suggestions:

You might want to define an explict bandwidth allocation for your IP Prec 4 class.

You might want to stop using (implicit) FQ within class-default.

You might need to tune your shapers, either smaller Tc and/or define less bandwidth (to allow for L2 header bandwidth).

Questions:

PE4 shows 1 Mbps Vbr-rt and CE4 shows 1 Mbps frame-relay? If these two interconnect you need to allow for the bandwidth difference between ATM and frame-relay (ATM uses more bandwidth because of cell header overhead and cell padding). Also, what's the max bandwidth of the frame-relay link since w/o a shaper (I believe) it's limited by the physical interface not CIR setting.