introduction to multimedia1 multimedia network. zreference: guojun lu, “communication &...

Introduction to Multimedia 1

Multimedia Network

Reference:Guojun Lu, “Communication & Computing for Distributing Multimedia Systems”Read Chapter page 124-150

Network Characteristics

1. The network bandwidth should be 1. The network bandwidth should be sufficiently high to support many sufficiently high to support many applications at the same time;applications at the same time;

2. Network resources should be shared 2. Network resources should be shared efficiently among applications so that as efficiently among applications so that as many applications as possible are many applications as possible are supported given certain resources;supported given certain resources;

3. The network should provide performance 3. The network should provide performance guarantees, if required, to applications;guarantees, if required, to applications;

4. The network should be scalable.4. The network should be scalable.

1. Network Speed vs Bandwidth

Sufficient BW for supporting multimedia communications.

Speed vs BW the time intervals representing each data bit “high-bandwidth network” instead of “high- speed network”

Network speed is determined by the physical transmission medium used, protocols, distance between intermediate nodes, and switching speed of intermediate nodes.

two different points of a network: user access pointuser access point

the bandwidth at this point is the bandwidth at this point is called user access bandwidth.called user access bandwidth.

network access point,network access point,

the same amount of bandwidth the same amount of bandwidth should be available for both should be available for both directions.directions.

2. Efficient Sharing of NW Resources

Each user requires a large amount of bandwidth, If each user reserves a bandwidth equal to its peak bit rate,

some bandwidth is wasted when the output bit rate is not at the highest solution: bandwidth on demand or statistical multiplexing discuss: Circuit switching technology?

Synchronous time division multiplexing (STDM)?Packet-switching network?Packet size?Retransmission of lost data?

A better approach to reduce the effect of lost packets on playback quality is to use traffic priorities the most likely source of data loss is buffer shortage in network exchanges. discard the lower priority data first

3. Performance Guarantee

the network should guarantee that a packet can access the network within a specified time

when the packet is on the network, it should be delivered within a fixed amount of time

when the performance is not guaranteed? transmission medium access control (MAC) transmission medium access control (MAC)

protocols?protocols? the network switches?the network switches?

Some techniques:1.Characteristics of different types of traffic should be

determined in terms of peak data rate, average data rate, bursty intervals, delay, and delay jitter requirements.

2. Network access time should be guaranteed.3. Network resources (bandwidth and buffer queues)

should be managed efficiently so that as many applications as possible are supported with performance guarantees.

4. Network Scalability

Three types of scalability: Distance ? Bandwidth ? The number of users ?

5. Multicasting Capability

There is a common need to distribute a stream to multiple destinations. For example, in a videoconference

It is slow and wasteful for the source to send a copy of the data to each destination one by one. Why?

Circuit switched or Packet switched ?

The suitable network for MM communication

The individual access bandwidth The individual access bandwidth should be at least a few Mbps. The should be at least a few Mbps. The aggregate bandwidth should at aggregate bandwidth should at least be in the order of 100 Mbps at least be in the order of 100 Mbps at the local area, and higher for the local area, and higher for WANs.WANs.

The network should be based on The network should be based on packet-switched statistical packet-switched statistical multiplexing instead of dedicated multiplexing instead of dedicated circuits for efficient sharing of circuits for efficient sharing of network resources.network resources.

The network should provide The network should provide throughput, error rate, delay, and throughput, error rate, delay, and delay jitter guarantee to delay jitter guarantee to applications.applications.

The network should be scalable in The network should be scalable in terms of bandwidth, the number of terms of bandwidth, the number of users, and distance covered.users, and distance covered.

The network should have The network should have multicasting capability. It is easier multicasting capability. It is easier to implement multicasting in to implement multicasting in packet-switched networks than in packet-switched networks than in circuit switched networks.circuit switched networks.

The suitable network for MM communication

End-to-End QoS for End-to-End QoS for Multimedia CommunicationMultimedia Communication

ICE BREAKER…..

“When communications networks are fast enough, multimedia communication can be achieved.”

Is this statement correct? Why?

IntroductionChallenges

Operational Issue Performance/Bandwidth Robustness/Stability Technology Integration

Business Issue Grow traffic and revenue Differentiate Services Respond quickly to customer application and

service requirements Efficiently capture and utilize key information for

flexible billing, planning, and monitoring

IntroductionServices Issues-1

Network Providers Increase utilization More money for some traffic Differentiation vs. Competition

Users Broad spectrum of choices Services enable applications

Industry New business area

IntroductionService Issue-2

What new differentiated services can be offered?

How to handle bandwidth hug multimedia applications and mission-critical application at the same time

How to handle congestion?How to provide scaleable performance for

new services?How to bill new servicesHow to support new services

Internet QoSNew Definition - Good Service

Better than “Best Effort Service” Service Classes

Enough for “Customer’s Needs” Bandwidth, Capacity, Usage Pattern

Support for “Various Applications” Multimedia Applications Mission-Critical Applications Secure Transactions ...

Internet QoSEffective Use of Networks

Exponential Growth Rate World-Wide 120 million subscribers today, 350 million in 3

years Ever increasing dial access numbers & speeds Increasing leased line access speeds Backbone bandwidth doubles every 6 months

Increasing bandwidth is only “passive solutions” Effective use of bandwidth is “active solutions”. Eliminate “overhead and useless usage”.

End-to-End Multimedia System


Introduction

Why resource management?Resources in Multimedia SystemsRequirements of a multimedia systemQuality of Service

Layered Model QoS Parameters QoS Classes

Operations


Why resource management?

Limited capacity in digital distributed systems despite data compression and utilization of new technologies.

End-to-end nature of multimedia applicationsProcessing of continuous data requires the

cooperation of every hardware and software component along the data path.

Competition for resources exists in an integrated multimedia system.

Resources required at different levels of a distributed multimedia system

Application, system, device, network...


Window of insufficient resources

High-qualityaudio

Interactivevideo

Network fileaccess

1980 1990 2000

insufficient

insufficient Sufficient but scarce

abundant


Resources in Multimedia Systems

Application Level

System Level (Operating System and Communication System)

Multimedia Device Level Network Level

Network host interface(bandwidth)

Multimedia devices(e.g. video/audio device)

CP

U

Mem

ory

Layered Partition of a Multimedia System wrt to required resources and processing/communication services


Resources - Classification

Resourceis a system entity required by tasks for manipulating

information.

Types of resourcesActive vs. passive resources -

• e.g. CPU is an active resource, main memory is a passive resource

Shared vs. exclusive resources• CPU is a shared resource, video board is an exclusive

resourceSingle vs. multiple resources

• In a uniprocessor the CPU is a single resource whereas in a multiprocessor system the CPU is a multiple resource.


Requirements

High level resource managementResource Reservation/Allocation/Adaptation

Process ManagementReal-time processing of continuous dataCommunication and Synchronization between

processes Memory Management

Guaranteed timing delay and efficient data manipulation functions

File System ManagementTransparent and guaranteed continuous retrieval of

audio/video. Device Management

Integration of audio/video devices


Huge Real-time requirement

Real-time system• Correctness of a computation depends not only on

obtaining the right results, but also upon providing the result on time.

Real-time process• is a process which delivers the results of processing in a

given time-span.

Real-time applicationsTemperature control in a chemical plant

• Driven by interrupts from an external device that occur at regular and unpredictable intervals.

Control of a flight simulator• scheduling of commands that execute at periodic

intervals. This scheduling is performed by a timer service which the application requests from its OS.


Deadlines

Deadlines represent the latest acceptable time for the presentation of a processing result.

Types of deadlinesSoft Deadline - In some cases the deadlines are

missed, but• not too many deadlines missed and• deadlines not missed by much.• E.g. flight arrival/departure

Hard Deadline • deadlines should never be violated since violations

imply system failure.• E.g. deadline failures in nuclear power plants, robotic

arm failures in process control systems


Requirements for Real-time operating systems

Multi-tasking capabilities• Real time application is divided into multiple tasks.• Helps keep CPU busy and ensures that processing of one

event is not blocked because the application is waiting for a different event.

Short interrupt latencyInterrupt latency

• is the time interval between a hardware device generating an interrupt and execution of the first instruction of the software interrupt handler.

• Interrupt Latency = hardware delay to get interrupt signal to processor + time to complete current instruction + time executing system code in preparation for transferring execution to the device’s interrupt handler.


Real-time operating system requirements

Fast context switchContext switch time (Dispatch latency)

• time between the OS recognizing that the awaited event has arrived and the beginning of execution of the waiting task.

Control of memory managementVirtual Memory vs. Real Memory

• A OS with VM support that aims to support real-time programming must provide a way for a task to lock its code and data into real memory so that it can guarantee predictable response to an interrupt.

Proper scheduling• OS must provide a facility to properly schedule time-

constrained tasks.


Real-time Operating System requirements

Fine granularity timer servicesneed access to fine granularity interval timesMillisecond resolution is a bare minimum,

microsecond resolution is required in some cases.Accurate time-of-day services

Rich set of intertask communication mechanisms.

Message queues, shared memory, Synchronization - semaphoresevent flags


Real-time and Multimedia

Audio and Videorepresent data streams of periodically changing

values.Semantics of multimedia information is dependant on

timely delivery.

Differences b/w real-time requirements for traditional real-time systems and multimedia systems

Fault tolerance and securitysoft deadline vs hard deadlineperiodic behavior vs. random behaviorbandwidth requirements


Resource management in networked multimedia systems

Application

System

Network

Application

System

Network

Network ResourceManager

Res

ourc

e M

anag

emen

t

Res

ourc

e M

anag

emen

t

End-SystemEnd-System

Switch

Qos Definition-1

“Quality of service represents the set of those quantitative and qualitative characteristics

of a distributed multimedia system necessary to achieve the required

functionality of an application.”

QoS Definition-2

“QOS is a quantitative and qualitative specification of an application’s

requirement, which a multimedia system should satisfy in order to

achieve desired application quality.”

Conceptual Model

Sample QoS Parameters


Quality of Service

Multimedia systems consist of a set of services These services may have different

requirements timeliness, delay, jitter, accuracy, performance etc.

These requirements are specified using QoS parameters.

Examples of QoS parametersAudio service

• sample rate of 8000 samples/sec, sample resolution of 8bits/sample

Network service• throughput of 100Mbps, connection setup time of 50ms


QoS concept (continued

QoS originated in the networking service domainprovided a specification of how good the offered

network services are (ISO standard definition).

Layered model of QoSextended QoS concepts to include not only networking

services, but also OS services and services for the provision of end-to-end QoS to the human user.

Services may be performed on different objectsmedia sources, sinks, connections, tasksQoS specification characterizes the service objects.Various definitions of QoS parameters

• real-time channels, streams, sessions, media...


Layered Model for QoS

User

NetworkMM devices

System

Application

(Network QoS)(Device QoS)

(System QoS)

(Application QoS)

(Perceptual QoS)

(Operating and Communication System)


Application QoS parameter examples

Application Qos

Media Quality …... Media Relations

Intraframe

Media Characteristics

Interframe

Component Spec

Name

Size

Rate

Importance

Loss Rate

Transmission Characteristics

Sample Size

Sample Rate

Compression

End-to-end Delay

Sample Loss Rate

Importance

Cost

Synchronization Skew

Integration

Communication

Conversion


System QoS parameter examples

System Qos

Application Subsystem Network Subsystem

Tasks per Connection Spec

Task Scheduler

Priority

Duration

Period

Task Ordering

Time Begin

Time Deadline

Space Requirements

Tasks per Medium Spec

Task Scheduler

Priority

Duration

Period

Task Ordering

Time Begin

Time Deadline

Space Requirements


Network QoS parameter examples

Network Qos for a connection

Throughput Spec Performance Spec Connection ID

Traffic Spec

Packet Size

Throughput

Burstiness

Packet Loss Rate

Intermediate Delay

Packet End-to-end Delay

Ordering

Error Control

Fragment/Reassembly

Communication Type

Cost

Priority


QoS Classes

QoS Service Classes determinereliability of offered QoSutilization of resources

Guaranteed Service ClassQoS guarantees are provided based on deterministic

and statistical QoS parameter values. Predictive Service Class

QoS parameter values are estimated and based on the past behavior of the service

Best-effort Service ClassNo guarantees or only partial guarantees providedNo QoS parameters are specified or some minimal

bounds are given.


Resource Allocation Tradeoffs in QoS Classes

appl2

unused

unused

Needs of appl 1

Needs of appl 2

Needs of appl 1conflict

Reserved for appl 1

Reserved for appl 2

Reserved for appl 1

Reserved for appl 2


Deterministic QoS parameter values

Single valueQoS1 - average value (QoS_ave), contractual value,

threshold value, target value.

Pair of values<QoS1,QoS2> -

• QoS1 - required value • QoS2 - desired value • <QoS_ave, QoS_peak>,<QoS_min,QoS_max> etc.

Triple of values<Qos1,Qos2,Qos3>

• QoS1 - best value, Qos2 = average value, QoS3 = worst value

• e.g. <BW_peak,BW_ave,BW_min>,similarly jitter….

Required value Desired value

Acceptable valueIncreasingquality


Guaranteed QoS

Need to provide • 100% guarantees for QoS values (hard guarantees) or • very close to 100% guarantees (soft guarantees)

Current QoS calculations and resource allocations are based on

• hard upper bounds for imposed workloads• worst case assumptions about system behavior

Advantages• QoS guarantees are satisfied even in the worst possible

case, hence high reliabilityDisadvantages

• Over-reservation of resource capacities, hence needless rejection of reservation requests.

• Qos values may in reality require softer bounds and significantly less resources than calculated hard bounds.


Predictive QoS parameters

AverageCan utilize QoS values in the past (QoS1,…QoSi) and

compute the average from 1 to i values. The desired QoS_bound at step K>I will be

• QoS_K = 1/i * QoS_j Maximum Value

Can utilize QoS values in the past (QoS1,…QoSi) and take maximum value as the desired QoS bound

• QoS_K = max QoS_i Minimum Value

Can utilize QoS values in the past (QoS1,…QoSi) and take minimum value as the desired QoS bound

• QoS_K = min QoS_i

j=1

i

i=1,..i

i=1,..i

i=1,..i


Best Effort QoS

No Qos bounds or possible soft QoS bounds Possible calculation of performance one might

get is based on• average case with stochastic description of imposed

workload• average case with stochastic assumptions about the

system behavior.

Advantages• resource capacities can be statistically multiplexed,

hence more processing requests can be granted.

Disadvantages• QoS may be temporarily violated, hence the service

guarantees are not reliable.


Relation between QoS and resources

Phase 1

Operations on QoSUser QoSrequirements

Resource reservation/allocation

QoS guarantees to user

Phase 2

QoS enforcement by resource control/adaptation

multimediaprocessing


QoS Operations

QoS Translation must be bidirectional

human interface (user QoS) - application QoSapplication QoS - system QoSsystem QoS - network QoS

Media Scaling (reverse translation) Transparent Scaling Non-transparent Scaling


Scaling - Examples

Audiotransparent scaling is difficult, non-transparent

scaling should be used.

VideoTemporal ScalingSpatial ScalingFrequency Scaling (reduction of DCT coefficients)Amplitude Scaling (reduction of color depth)Color Space Scaling (reduction of the number of

entries in the color space)


QoS Negotiation

User (Caller)

Application (Caller)

System (Caller)

.

.

.

User (Callee)

Application (Callee)

System (Callee)

.

.

.

Caller to Callee

Peer to Peer

Service User to Service Provider

Layer to Layer

Overview of the QoS Negotiation Operation


Bilateral Peer to Peer Negotiation

Caller Callee

Service Provider

Peer to Peer

Negotiation

Connect indication

Connect request

Connect response

Connect confirm

Requested QoS value

Changed QoS value

t_1 t_2 t_3t_4


Triangular QoS Negotiation

Caller Callee

Service Provider

Peer to Peer

Negotiation

Connect indication

Connect request

Connect response

Connect confirm

Requested QoS value

Changed QoS value

t_1 t_2 t_3t_4

Changed QoS value


Qos Negotiation in the Layered Model

Application-to-networkcommunication

Peer-to-Peercommunication

Tuning of Application QoS parameters

Negotiate/RenegotiateApplication QoS parameters

Negotiate/RenegotiateNetwork QoS parameters

QoS Translator(Bidirectional translation)

Accept/modify Reject


Resource Management Operations

Resource Management consists of• resource managers• resource management protocols and services

Operations performed by resource managers to provide QoS

Establishment Phase• schedulable units utilizing shared resources must be

admitted and resources must be reserved and allocated.

Enforcement Phase• Reserved and allocated resources must be provided,

enforced and possibly adapted according to QoS requirements during the lifetime of the applications.


QoS Establishment

The following operations must be performedQoS to resource mapping

• Need translation profilesResource admission

• Need admission tests to check availability of resourcesResource reservation

• must be done along the path from initiator to initiateeResource allocation

• done along the path from initiatee to initiator

Need a continuous media resource model• need to know how applications will use resources• model chain of resources during the traversal of

multimedia messages along the end-to-end path in a distributed multimedia system


Admission Tests

Task schedulability tests for CPU Packet schedulability tests for sharing host

interfaces, switches• needed to provide delay and jitter reliability

guarantees.

Spatial Tests for Buffer Allocation• needed for delay and reliability guarantees

Link Bandwidth tests• throughput guarantees


Resource reservation and allocation

2 types of reservationsPessimistic Approach

• worst case reservation of resourcesOptimistic Approach

• average case reservation of resources

For resource reservation, we need• Resource Table - information about managed resources• Reservation Table - current reserved resources• Reservation Function - maps QoS to resources and

operates on the reservation table.

Reservation Styles• Sender initiated• receiver initiated


QoS Enforcement

Resource Schedulinge.g. rate monotonic scheduling

Rate Control - Traffic shapinge.g. leaky bucket

End-to-end Error Controle.g. forward error correction

Flow ControlOpen Loop Flow Control (no feedback)Closed Loop flow control (with feedback)

Flow synchronization


QoS Management Issues

Resource and QoS Monitoringshould be flexible, variables should be optional and

monitoring can be turned on and off• User mode monitoring• Network mode monitoring

QoS Maintenance• compares monitored QoS with contract QoS

QoS Degradation• issues a QoS indication

QoS Signaling• specifies interval over which QoS should be monitored

and user informed about the delivered performance QoS Scalability

• includes QoS filtering and adaptation


QoS Renegotiation

RenegotiationEntity

Network Management

Application Management ( User Interface) Application QoSparameters

Network QoSparameters

Signal “Change Application QoS”

Signal “Change Network QoS”


QoS Resource Adaptation

As a result of a renegotiation request (request for change in quality)

adaptation in QoS and adaptation in resource allocation must occur.

Renegotiation request can come from• User• Host system• Network

Kinds of adaptationNetwork Adaptation (e.g dynamic rerouting

mechanism)Source Adaptation (e.g. temporal scaling with

feedback)


Resource Deallocation

Free up resources used by requestTear down process:

Sender initiated closingReceiver initiated closing


QoS Provisioning - Current State

IncompletenessInterfaces not QoS configurableOnly small sets of facilities to support MM flows exist

Lack of mechanisms to support QoS guarantees

Need research in distributed control, monitoring, adaptation, maintenance of QoS mechanisms

Lack of overall frameworksQoS frameworks for heterogeneous environments


Design Principles for QoS

Integration • QoS should be configurable, predictable and

maintainable over all layers.

Separation • separation of transfer, control and management

Transparency • applications must be shielded from the complexity of

underlying QoS specification.

Asynchronous Resource Management• supports the functionality division between

architectural modules.

Performance• try to avoid multiplexing, layer processing, use

hardware...

QoS Mechanism

QOS can be guaranteed only when: sufficient resources are available and proper scheduling of processes is implemented

How to provide QoS guarantees??

Preventing overload requires admission control, and preventing applications using more resources than what is allocated requires policing mechanisms.

The four important subsystems of End-to-End Multimedia systems are networks, transport protocols, end-system architecture, and multimedia server.

introduction to multimedia1 multimedia network. zreference: guojun lu, “communication &...

Documents

network bandwidth

network resources bandwidth

highbandwidth network

network switches

suitable network

network exchanges

network characteristics1

network access time