內(nèi)容運(yùn)營商系列-qos22s07l p

Thesemechanisms

must

classify

packets

beforepolicing

or

shathe

traffic

rate.Traffic

sha

queues

excess

packets

to

stay

within

the

desiredtrafficrate.Traffic

policing

typically

drops

or

marksexcess

traffic

to

stay

within

atraffic

ra

imit.Ra imit

file-sharing

application

traffic

to

56kbps.Do

not

ra imit

trafficfrom

mission-critical

server.Central

to

remote

site

speed

mismatchRemote

to

central

site

over

subscriptionBoth

situations

result

in

buffering

and

indelayed

or

dropped

packetsing

and

outgoingdirectionsOut-of-profile

packets

are

droppedDrop causes

TCPretransmitsSupports

packet

marking

orre-markingLess

buffer

usage

(sharequires

an

additional

shaqueuingsystem)Outgoing

direction

onlyOut-of-profile

packets

arequeueduntil

a

buffer

gets

fullBuffering

minimizes

TCPretransmitsMarking

or

re-marking

notsupportedShaFramsupports

interaction

withRelay

congestion

indicationIf

sufficient

tokens

are

available

(conform

action):Tokens

equivalent

to

the

packet

size

are

removed

from

thebucket.The

packet

is

transmitted.If

sufficient

tokens

are

NOT

available

(exceed

action):–

Drop

(or

mark)

the

packetBc

is

normal

burst

sizeTc

isthe

time

intervalCIR

is

the

committed

information

rateCIR

=

Bc/

Tc的token注入方法是：每當(dāng)一個(gè)數(shù)據(jù)包，就注入tokenEvery

time

a

packet

is d,

CB

policing

puts

sometokens

back

intothe

Bucket.

The

number

of

tokensplaced

into注入數(shù)量如下：(Current_packet_arrival_time

–

Previous_packet_arrival_time)

*_rate/8Be:

Excess

burst

sizeTc:

Tokens

in

Bc

bucketTe:

Tokens

in

Be

bucketThe

return

value

is

conform

or

exceed

or

violateTc:

Tokens

in

CIR

bucketTp:

Tokens

in

PIR

bucketEnforce

traffic

policing

according

to

two

separate

rates:Committed

Information

RatePeak

Information

RateTwo-rate r

marks

packets

as

either

conforming,exceeding,

or

violating

a

specified

rate.If

(B

>

Tp),

the

packet

is

marked

as

violating

the

specifiedrate

都不減少token數(shù)量If

(B

>

Tc),

the

packet

is

marked

as

exceeding

the

specifiedrate,

and

the

Tp

token

bucket

is

updated

asTp

=

Tp

–

B此時(shí)有個(gè)問題，下次再來一個(gè)數(shù)據(jù)包，可能Tp<Tc；If

the

packet

is

marked

as

conforming

to

the

specified

rate,and

both

token

buckets(Tc

and

Tp)

are

updated

as

Tp=Tp–B

and Tc=Tc–B.

同時(shí)減少token的數(shù)量單速率配置方法：以下4個(gè)配置命令含義一樣80000

1000

1000cir

80000

1000

1000cir

80000

bc

1000

be

1000rate

80000

bps

burst

1000

bytes

peak-burst

1000

bytes雙速率配置方法：policy-map

yeslab7class

p1cir

8000

bc

1500

pir

10000

be

5000conform-action

transmitexceed-action

set-prec-transmit

1violate-action

set-prec-transmit

1r2#show

policy-map

interfaces2/0Serial2/0Service-policyoutput:

yeslab7Class-map:

p1

(match-all)15

packets,

1052

bytes5

minute

offered

rate

0

bps,

drop

rate

0

bpsMatch:

any:cir

8000

bps,

bc

1500

bytespir

10000

bps,

be

5000

bytesconformed

1

packets,

48

bytes;

actions:transmitexceeded

0

packets,

0bytes;

actions:set-prec-transmit1violated

0

packets,

0

bytes;

actions:set-prec-transmit1conformed

0bps,

exceed

0

bps,

violate0

bps每隔Tc，往桶里注入Bc個(gè)token，是 注入，不是勻速注入。然后sh r

gateopen，什么時(shí)候關(guān)閉呢？在桶里沒有token的時(shí)候就關(guān)閉了，然后在下一個(gè)Tc開始的時(shí)候在打開并

注入Bc個(gè)token發(fā)送單個(gè)數(shù)據(jù)包的速率等于link的物理層速率通過減少發(fā)送時(shí)間來降低速率如上圖：link速率128K，限制到64k，就只能用一半的時(shí)間用來發(fā)送數(shù)據(jù)流，設(shè)置TC的值為125ms，因此每次發(fā)送數(shù)據(jù)包的時(shí)間就為62.5ms，Bc＝62.5ms*128Kbps＝8000bit填充token的方法：每125ms的開始就填充Bc個(gè)tokenWhen

configuring

sha

,you

typically

configure

the

sha

rateand

optionally

theBc（典型假設(shè)TC為4ms）.Ifyou

configure

bothvalues,IOS

changes

the

Tc

so

that

the

formula

is

met配置舉例：sh average

percent

10

125ms紅色為CIR的值，用帶寬比例表示紫色為TC的值，單位為ms自動(dòng)計(jì)算出的BC值為：19300Service-policy

output:

yeslabClass-map:

p1

(match-all)138

packets,

37338

bytes5

minute

offeredrate

0

bps,

drop

rate

0

bpsMatch:

anyQueueingqueue

limit

64packets(queue

depth/total

drops/no-buffer

drops)

0/0/0(pkts

output/bytes

output)

0/0sh (average)

cir

154400,

bc

19300,

be

19300sh rate

154400配置舉例：sh average

1000000

12000紅色表示CIR紫色表示BCTC值將會(huì)自動(dòng)計(jì)算：12000/1000000＝120msService-policy

output:

yeslabClass-map:p1

(match-all)298

packets,

113530

bytes5

minute

offeredrate

0

bps,

drop

rate

0

bpsMatch:

anyQueueingqueue

limit

64packets(queue

depth/total

drops/no-buffer

drops)

0/0/0(pkts

output/bytes

output)

160/76614sh (average)

cir

1000000,

bc

12000,

be

12000sh rate

1000000對(duì)sha

的單桶單速率無BE，桶的大小為BC，如果一個(gè)數(shù)據(jù)包來，token夠用，轉(zhuǎn)發(fā)，同時(shí)減少token相應(yīng)的數(shù)量；如果一個(gè)數(shù)據(jù)包來，token不夠，等下一個(gè)tc的注入的token。從宏觀效果來看，總體速率保持為CIR對(duì)sha

的單桶單速率有BE，桶的大小為BE+BC。該類型的sha

增加了對(duì)bursty流量的能力，但是從宏觀效果來看，總體速率還是保持為CIRpolicy-map

yeslab8class

p1sh average

1000000

12000

6000r2#show

policy-map

inter

s2/0Serial2/0Service-policy

output:

yeslab8Class-map:

p1

(match-all)38

packets,

2527

bytes5

minute

offered

rate

0

bps,

drop

rate

0

bpsMatch:

anyQueueingqueue

limit

64

packets(queue

depth/total

drops/no-buffer

drops)

0/0/0(pkts

output/bytes

output)

38/2588sh

(average)

cir

1000000,

bc

12000,

be

6000sh rate

1000000Traffic

sha and

policing

are

mechanisms

that

useclassification

to

limit

traffic

rate.Traffic

sha queues

excess

packets

to

stay

within

thecontractual

rate.

Traffic

policing

typically

drops

excess

trafficto

stay

within

the

limit;

alternatively,

it

can

re-mark,

then

sendexcess

traffic.Both

traffic

policing

and

sha

ensure

that

traffic

doesnotexceed

a

bandwidth

limit,

but

they

have

different

impacts

onthe

traffic.Traffic

rate

is

metered

using

a

token

bucket

mathematicalmodel.With

traffic

policing,

new

tokens

are

added

into

the

tokenbucket

based

on

the

CIR.Class-based

traffic

policing

can

be

configured

to

supportexcess

bursting

capability.

With

excess

bursti