tu/etu/e eindhoven university of technology 6 content characteristics: video semantics (i) a typical...

60
TU/e eindhoven university of technology 1 Advanced Topics in Multi- Service Networks I Lecture 3: Traffic characteristics, admission control and Integrated Services

Upload: others

Post on 02-Jan-2021

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 1

Advanced Topics in Multi-Service Networks I

Lecture 3: Traffic characteristics, admission control and Integrated Services

Page 2: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 2

Scheduling of isolated traffic flowsOverview and characteristic propertiesFCFS, Round Robin, GPS, WFQ, EDF

Queuing Theory RevisitedLittle’s Law, Balance EquationsM/M/1, M/M/n, M/D/1

Clever Queuing: Congestion ControlFlow control with TCPQueue Management to avoid congestion

Overview

Page 3: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 3

Traffic characterisationTraffic classes and their characteristicsDeterministic Traffic DescriptorsStochastic Traffic Descriptors

Admission ControlOverviewParameter vs. measurement based AC

Integrated Services FrameworkResource reservation with RSVP and variantsScalability limitations and other problems

Overview

Page 4: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 4

Correlation Between Content and Traffic (I)

Image: 256x256 grey scale

JPEG Q=5, 0.25 bits/pixel (2080 bytes)

JPEG Q=20, 0.54 bits/pixel (4442 bytes)

original GIF(54749 bytes)

Page 5: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 5

Correlation Between Content and Traffic (II)

Image: 512 x 512 colour

JPEG, Q=5, 0.22 bits/pixel (7128 bytes)

original GIF (202749 bytes)

JPEG, Q=20, 0.43 bits/pixel (13984 bytes)

Page 6: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 6

Content Characteristics: Video Semantics (I)

A typical video sequence is usually described as a collection of independent video shots called scenes.Each scene is an ordered set of video frames depicting a real-time continuous actionExcept from static real-world scenes, a typical video sequence consists of various artificial effects (e.g. camera movement, zooming, picture in picture, graphics etc.)The assumption of scene independence does not seem to hold in most video sequencesThe scenes are in fact correlated. There is a low probability of having a short scene while the probability of having long ones decades relatively slow

Page 7: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 7

Video Compression Techniques

Bandwidth requirements for Uncompressed video: 100-240 MbpsUncompressed HDTV streams: around 1Gbps

Compression is achieved by removing redundancyRedundancy types in video signal:

within a video frame (intra-frame): spatialbetween frames in close proximity (inter-frame) : temporal

Compression methods:Lossless or lossy (w/o exact recovery)Symmetric or asymmetric (same computational effort in coder and decoder or not)

MPEG-1 Quality similar to that of a VHSBit-rate approximately 1.2 Mbps

MPEG-2Broadcast quality videoBit-rate: 4-6 MbpsWide range for rate and resolution

Page 8: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 8

Characteristics of Compressed Video

Statistical behavior is described by using histograms and correlation functionsCorrelations arise as a consequence of visual similarities between consecutive images (or parts of images) in video streamsCompression results in a reduction in these correlationsCompressed stream still contains considerable amount of correlations Correlation along with different information content of each frame leads to burstiness

Page 9: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 9

Video Semantics (I)

Page 10: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 10

Burstiness

One simple definition: Ratio between peak and average bit rate (PAR)Burstiness indicates the presence of non-negligible positive correlations between cell inter-arrival times

Page 11: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 11

Other Bit Rate Burstiness Measures

Bit-rate distributionAverage bit-ratePeak bit-rateVariance

Autocorrelation FunctionExpresses temporal variations

Coefficient of VariationUncorrelated streams have fixed CoVUseful when measuring multiplexing characteristics when VBR signals are buffered and statistically multiplexed

Scene duration distributionAverage duration of peaks

Useful to estimate probability of buffer overflow

Page 12: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 12

Classifying Parameters for Traffic

Mean packet rate Variance of instantaneous packet rateCorrelation of instantaneous packet rateOthers (e.g. higher order moments and cumulants)?Which classifying parameters are important?

Page 13: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 13

Model ?

Loss probability, delay

Unfinished work distribution

mean

variance

autocovarianceOther? Other?

All performance techniques must make some assumptions regarding traffic or workload

Queuing models: arrival and service processesSimulation: traffic generatorsExperimental: work load generators

Assumptions are captured in traffic models

Traffic Models

Page 14: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 14

Developing Traffic Models

What is the basis for traffic models?Select from a constrained set

Analytical techniques limit the types of models´that can be utilized, e.g., MarkovianSpecific model may be based more on “tradition” than realityImportant to understand limitations and inaccuracies of such traffic assumptions

Analysis of trace dataA trace is captured from a “live” networkLive traffic may not exist, e.g., for a new network or applicationHow do we know our sample is typical or large enough?

Use insight into applicationsModels based on expert knowledge of the system

Page 15: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 15

Continuous Time Traffic Source Models

Events (arrivals and departures) can occur at any instant of time in continuous time modelsPoisson Process (PP or M)

Appropriate if there is a large number of independent users and each generates Poisson traffic and no source dominatesPoisson arrival process basis of standard queuing models (M/M/1,M/D/1, M/M/m/m, etc.)

Generally Modulated Poisson Process (GMPP):A Poisson arrival process, but with time-varying arrival rate,The arrival rate is determined by the modulating process

Markov Modulated Poisson Process (MMPP)A GMPP where the modulating process has a Markov chain

)(tλ)(tλ

Page 16: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 16

Rate is controlled by the modulating process (two-state Markov chain here, but could be more states)The arrival process is Poisson for given rate

Switched Poisson Process (SPP)Special case of a MMPP where the modulating process is a two-state Markov chainRate is when system is in state 0 and rate is when system is in state 1

Interrupted Poisson Process (IPP)

1λ0λ

Markov-Modulated Poisson Process

Special case of a SPP where rate and rate 00 =λ 01 >λ

Page 17: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 17

Discrete Time Traffic Source Models

Events (arrivals and departures) occur at slot timesOften assumes that one packet (cell) can be transmitted in a single slot, as in ATM

Deterministic Process (DP)Events occur at fixed multiples of the basic slot, e.g. transmission of an ATM cell requires one slot

Bernoulli Process (BP)Discrete time counterpart to the Poisson ProcessTractable first-cut model for discrete time systems

Page 18: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 18

Stochastic Models – Long Range Dependence

In many processes (e.g., Poisson), the autocovariance rapidly decays with t

Short-range dependent processes, covariance decays at least exponentially

A long-range dependent process has a hyperbolically decaying autocovariance

Long-range dependence reflects the existence of clustering and bursty characteristics at all time scales in self-similar processes

Self-similarity:A phenomenon that is self-similar looks or behaves the same when viewed at different degrees of “magnification” or different scales on a dimension (time or space)We are concerned with time series and stochastic processes that exhibit self-similarity with respect to time

Page 19: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 19

A self-similarstructure containssmallerreplicas of itself at all scales

Self-similarity Conditions:

[ ] [ ]HaatXEtXE )()( =

[ ] [ ]H

XXXX a

atXRtXR 2)()( =

[ ] [ ]Ha

atXVartXVar 2)()( =

Mean

Variance

Autocorrelation

15.0 ≤≤ H

Hurst Parameter:

H = 0.5 :No Self-similarity

Self Similarity

Page 20: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 20

Is Ethernet Traffic Self Similar

Page 21: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 21

Self-Similarity of Ethernet Traffic

Estimated Hurst parameter of H = 0.9The higher the load, the higher HAggregating several streams does not remove self-similarity

Failure of Poisson ModelingInstead Superposition of many Pareto-like ON/OFF sources

Pareto distribution exhibits long-range dependenceEach source alternates between “ON” and “OFF” statesIn ON, a burst of packets is transmitted, an OFF period is an idle period

Page 22: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 22

Pareto Distribution

Page 23: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 23

mA(t), l D(t)

Model of FIFO router queue

Deterministic upper/lower bound on trafficQ(t)

D(t)

A(t)

time

Q(t)d(t)

Cumulativenumber of bits

Facilitates computation of deterministic upper/lower bounds on network performance

)()()( stsAtA −≤− α

Deterministic Traffic Model

Page 24: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 24

Peak Rate is most conservative traffic envelope

tpt =)(α

time t

data

A(t)

0

b

time t

data

A(t)

0

b

p

Leaky Bucket is a frequently deployed traffic envelope (for policing)

btt += ρα )(

ρ

Deterministic Models – Traffic Envelope (I)

Page 25: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 25

Dual Leaky Bucket/Token Bucket as straight forward and computationally tractable extension

[ ]bttpt += ρα ,min)(

data

A(t)

0 time t

data

A(t)

0 time t

bp

N-segment piecewise linear service curve as compromise between accuracy and computational complexity

[ ]iiibtt += ρα min)(

Deterministic Models – Traffic Envelope (II)

Page 26: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 26

Leaky Bucket

data

B

rate, L

Two parameters:B: bucket size [Bytes]L: leak rate [B/s or b/s]

Data pours into the bucket and is leaked outB/L is maximum latency at transmissionTraffic always constrained to rate L

Page 27: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 27

Three parameters:b: bucket size [B]r: bucket rate [B/s or b/s]p: peak rate [B/s or b/s]

Bucket fills with tokens at rate r, starts fullPresence of tokens allow data transmissionBurst allowed at rate pdata sent < rt + b

data

tokens, rate r

b

peak rate, p

Token Bucket

Page 28: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 28

Traffic characterisationTraffic classes and their characteristicsDeterministic Traffic DescriptorsStochastic Traffic Descriptors

Admission ControlOverviewParameter vs. measurement based AC

Integrated Services FrameworkResource reservation with RSVP and variantsScalability limitations and other problems

Overview

Page 29: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 29

Single Node Admission Control

Node receives a service request and checks:Buffer sizeAvailable capacity (depending on scheduling discipline)

Possibly local computation of performanceReject or admit flow

Page 30: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 30

End-to-end Admission Control

Parameter based Admission Control Deterministic Traffic ModelStateful Network (allocated resources known)

Measurement Based Admission ControlMeasurement of recent traffic load historyMeasured load + new flow vs network capacity

Network Probing (Measurement Variant) Alleviates us from continuous traffic measurementSending probing packets to obtain instantaneous network load indication

Page 31: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 31

Measurements vs Parameter Admission Control

Parameter based Admission Control: Facilitates provision of deterministic performance guaranteesDecision based on worst case boundsTypically, low network utilization

Measurement based Admission Control:No deterministic QoS guarantees, statistical guarantees onlyDecision based on traffic measurementsHigher utilization than parameter based

Page 32: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 32

Measurement Based Admission Control

Many different varieties of MBACs:Some based on “solid” math models (eg, theory of large deviations) Others “ad hoc” (no theory underpinning)Different load estimations: from simple point estimate, to exponential averaging , combined mean and variance measurements, etc

Two Components:Network load measurements (on aggregate rather than per flow)Admission control decision based on load measurement and trafficprofile indicated in service request

Page 33: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 33

Probe Based Admission Control

Variant of Measurement Based Admission Control, where:Not the (aggregate) traffic load is measured at the ingress, butThe load in the network is measured by means of probing packets

No need for stateful network management, e.g. no link state advertisementsAdmission decision based on measured network load and indicated traffic profile of new serviceCriticalities:

Probing measures instantaneous load, in case of high (aggregate)traffic dynamics load measure may not be reasonableDanger of cheating: probing packets may be explicitly dropped bydomains on purposeBottleneck link sometimes hard to determine

Page 34: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 34

Available Bandwidth

Probing

Active Probing

Packet Train

Packet Pair

Trains of Packet Pairs

Variable Packet Size

Probing Taxonomy

Page 35: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 35

General Definitionn : number of hops in a path

: capacity of linki: avail-bw of the path at time t: utilization of linki

Link1=2Mbps=20%

Link2=1Mbps=30%

Link3=3Mbps=50%

A(2) = C1 x (1-u1) = 1.6MbpsA(2) = C2 x (1-u2) = 0.7MbpsA(2) = C3 x (1-u3) = 1.5Mbps

Sender

Receiver

( ) ( ) ( ) ( )( )1...

min 1t ti ii n

A n c u n=

⎡ ⎤= −⎣ ⎦ic

iu

1u1c

2c2u 3c

3u

( )tA

A(2) = C2 x (1-u2) = 0.7Mbps

Probe Capacity Measurement

Page 36: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 36

P2 P1

Gap (time) gi

P2 P1

go

router

P2 P1 : back-to-back packet pair

Cross traffic packets

gi: initial gap (time between first bit of P1 and P2 when they enter the router)

go: output gap (time betweenfirst bit of P1 and P2 when they leave the router)

IGI - Initial Gap Increasing

Benefit: Reflect whether the link is bursting or not

Tradeoff: The length of giToo short: Cause probe overflow Too long: Underestimate the available bandwidth

Page 37: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 37

Traffic characterisationTraffic classes and their characteristicsDeterministic Traffic DescriptorsStochastic Traffic Descriptors

Admission ControlOverviewParameter vs. measurement based AC

Integrated Services FrameworkResource reservation with RSVP and variantsScalability limitations and other problems

Overview

Page 38: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 38

IETF Integrated Services

Architecture for providing QOS guarantees in IP networks for individual application sessionsResource reservation: routers maintain state info of allocated resources and performance requirementsAdmit/deny new call setup requests:

Question: can newly arriving flow be admittedwith performance guarantees while not violatedQoS guarantees made to already admitted flows?

Page 39: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 39

IntServ: QoS guarantee scenario

Resource reservationcall setup, signalingtraffic, QoS declarationper-element admission control

QoS-sensitive scheduling (e.g., WFQ)

request/reply

Page 40: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 40

Call Admission

Arriving session must :Declare its QOS requirement

R-spec: defines the QOS being requestedCharacterize traffic it will send into network

T-spec: defines traffic characteristicsSignaling protocol: needed to carry R-spec and T-spec to routers (where reservation is required)

RSVP

Page 41: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 41

IntServ Specifications

Page 42: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 42

Guaranteed service:worst case traffic arrival: leaky-bucket-policed source simple (mathematically provable) bound on delay [Parekh 1992, Cruz 1988]

Controlled load service:"a quality of service closely approximating the QoS that same flow would receive from an unloaded network element."

WFQ

token rate, r

bucket size, bper-flowrate, R

D = b/Rmax

arrivingtraffic

T-spec for traffic specification

IntServ QoS: Service Models [rfc2211, rfc2212]

Page 43: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 43

Resource Reservation Signalling

RSVP, Resource-Reservation ProtocolRFC 2205Enable resources to be reserved for a given session in priorThe most complex, and closest to circuit emulation

Strong QoS guaranteesSignificant granularity of resource allocationSignificant feedback to applications

Two levels of serviceGuaranteed - as close as possible to circuit emulationControlled load – equivalent to the service in a best-effort network under no-load conditions

Page 44: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 44

Soft State

Routers keep state about reservation.Periodic messages refresh state.Non-refreshed state times out automatically.Alternative: Hard state

No periodic refresh messages.State is guaranteed to be there.State is kept till explicit removal.Why could there be a problem?

Properties of soft state:Adapts to changes in routes, sources, and receivers.Recovers from failuresCleans up state after receivers drop outs

Page 45: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 45

RSVP Signalling Flow (I)

Page 46: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 46

RSVP Signalling Flow (II)

The RSVP SENDER_TSPEC objectCarries the traffic specification (sender TSpec) generated by each data source within an RSVP sessionTransported unchanged through the network, and delivered to both intermediate nodes and receiving applications

The RSVP ADSPEC objectCarries information which is generated at either data sources or intermediate network elements,Flows downstream towards receivers, and may be used and updated inside the network before being delivered to receiving applications

The RSVP FLOWSPEC objectCarries reservation request (Receiver_TSpec and RSpec) information generated by data receiversFlows upstream towards data sourcesMay be used or updated at intermediate network elements before arriving at the sending application

Page 47: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 47

TSpec

A token bucket specificationbucket size, btoken rate, rthe packet is transmitted onward only if the number of tokens in the bucket is at least as large as the packet

peak rate, p>rmaximum packet size, Mminimum policed unit, m

All packets less than m bytes are considered to be m bytes

Page 48: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 48

Flowspec

An indication of the QoS control service requestedControlled-load service and Guaranteed service

For Controlled-load serviceSimply a Tspec

For Guaranteed serviceA Rate (R) term, the bandwidth required

R ≥ r, extra bandwidth will reduce queuing delaysA Slack (S) term

The difference between the desired delay and the delay that would be achieved if rate R were usedUsed to reduced the resource reserved

Page 49: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 49

AdSpec

PATH(TSpec)RESV(flowspec)The receiver to be informed about the network

The receiver does not request what the network cannot provide

The sender and routersIndicate their QoS capabilities; advertisingThe sender constructs an initial ADSpecEach router update the ADSpecAlso indicate that one or more routers RSVP-incapable

Page 50: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 50

Router Mechanisms

Page 51: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 51

Reservation Errors

A given resource reservation failsAn error message is returned

PathErr messages simply sent back to the senderResvErr messages are sent to a receiver

Only to the receiver whose request fails

Page 52: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 52

Guaranteed Service Specification

Tspec parameters are:P = peak rate of flow (bytes/sec)B = bucket depth (bytes)r = token bucket rate (bytes/sec)m = minimum policed unit (bytes)M = maximum datagram size (bytes)

Rspec parameters are:R = service rate (bytes/sec)

The rate that will be reserved at every routerS = slack term (micro seconds)

The maximum delay tolerancea router if does not have enough available bandwidth as requested by the receiver in R, then it reserves R’ that will cause increase in delay di such that di < S.

Page 53: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 53

Guaranteed Service Performance

Each router reserves bandwidth R and buffer space BThe delay bound will be: b/R + C/R + DThe loss bound will be: b + C + (D*R)Error terms C and D (may be included in rate-latency model):

C is rate dependent error factor measured in bytesD is rate independent error factor measured in micro-secondsC and D will be included in scheduler’s latency

Traffic Shaping at the source: The traffic must not exceed at any time T: M+min(pT, rT+b-M)

With peak rate p and maximum packet size M more accurate bound equations are:

( )+

+

⎥⎦

⎤⎢⎣

⎡−

−−

−++=

++−−−

=

δδ

δ

rpMbRrrbQ

RM

rpRRpMbD

)(])[(

Page 54: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 54

Reservation Styles

Three stylesWildcard/No filter – does not specify a particular senderFixed filter – sender explicitly specified for a reservation

Video conference

Dynamic filter – valid senders may be changed over time

Audio conference

Page 55: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 55

Example

S1 A

S2, S3

B

CR1

D R2R3

Senders S1, S1, S3Receivers R1, R2, R3Router interfaces A,B,C,C

Page 56: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 56

Wildcard Filter Reservation

R1, R2, and R3 want to reserve 4b, 3b, and 2b, respectively (b is given rate).

4b

3b

4b

4b

S1 A

S2, S3

B

CR1

D R2R3

Page 57: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 57

Fixed Filter Reservation

R1 wants to reserve 4b for S1 and 5b for S2.R2 wants to reserve 3b for S1 and b for S3.R3 wants to reserve b for S1.

S1:4bS2:5bS1:4b

S2:5bS3:b

S1 A

S2, S3

B

CR1

D R2R3S1:3b

S3:b

Page 58: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 58

Dynamic Filter Reservation

R1 wants to reserve b for S1 and S2 (shared).R2 wants to reserve 3b for S1 and S3 (shared).R3 wants to reserve 2b for S2.

(S1,S2):b

(S1,S2,S3):3b

S1:3b

(S2,S3):3b

S1 A

S2, S3

B

CR1

D R2R3

Page 59: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 59

ScalabilityEvery router needs to keep per flow RSVP state

It does not recognize internet domain architectureAutonomously administered domainsIt is a gross assumption that every domain will provide the same bounded service

Service definition is up to the service providersIt is business orientedConsistent end-to-end services difficult to achieve

IntServ Criticalities

Page 60: TU/eTU/e eindhoven university of technology 6 Content Characteristics: Video Semantics (I) A typical video sequence is usually described as a collection of independent video shots

TU/eeindhoven university of technology 60

Traffic characterisationSource Traffic ModelsDeterministic Traffic ModelsSelf Similarity

Admission ControlParameter based ACMeasurement based ACProbing based AC

IETF Integrated ServicesResource reservation and RSVPQoS ClassesCriticalities

Lecture Summary