Multimedia Applications

1
 Aims and Contents
• Aims
– Define multimedia applications
– Introduce technologies for delivering multimedia
applications in the Internet
– Raise the issues in delivering multimedia applications

• Contents
– Application characteristics and requirements
• Voice, audio, video; Performance requirements
– Voice over IP
– Streaming Stored Audio/Video
– IPTV

– (and Multimedia and QoS)

2 2
Application Characteristics and
Requirements

3
 Types of Multimedia Applications
• One-way (unidirectional) communications
– Listening to radio/music; viewing recorded or live video
– Also referred to as Streaming
• Stored audio/video
• Live audio/video
• Two-way (bidirectional) communications
– Voice calls, video-conferencing
– Also referred to as Interactive

• Sometimes multimedia applications are referred to as real-time

applications
– Rela-time communications: sender and receiver communicate
as if they were at the same physical location
• Very small delay and / or jitter between sender and
receiver is necessary
4
 Multimedia vs Other Applications
• “Other” Applications
– Web browsing, file download, email, database access, …
– Reliability is essential
– Large and/or varying delays can be tolerated

• Multimedia applications
– Delay-sensitive: large delays or jitter make the application un-
useable
– Loss-tolerant: if some data is lost in video or audio, the
application is still usable

5
 Voice Communications
• The human voice uses frequencies in the range of 100Hz to <10KHz
– Majority of voice communications is 300HZ to 3400Hz

6
 Voice Communications
• Analog voice data is converted to digital data using pulse
modulation techniques, e.g.
– Pulse Code Modulation (PCM), Delta Modulation

Levels

7
 Voice Communications
• How much information is needed in digital data to accurately
reproduce analog data at receiver?
– Sampling Rate: how often (in 1sec) is the analog data sampled?
• Units: Samples per second or Hertz (Hz)
• Nyquist’s Sampling Theorem tells us if we sample twice the
highest frequency signal component then can make a perfect
reproduction
• If highest frequency component is B, then sampling rate should be
2B
– Sample Size: how many different levels can a sample represent?
• Units: bits

– Low sample rates and small sample sizes can lead to poor
voice reproduction at the receiver

– High sample rates and large sample sizes require high

transmission data rates (bandwidth will increase)
8
 Voice Communications
1HZ sine wave

50 samples per second 50 samples per second 50 samples per second

100 levels 10 levels 4 levels

50 samples per second 10 samples per second 4 samples per second

10 levels 10 levels 10 levels
9
 Audio Communications

• Voice communications is a type of audio communications

• Non-voice audio communications: e.g., music
• Same concepts apply as for voice communications
– Music generally has a larger bandwidth than voice
– Users often desire higher quality output

10
 Video Communications
• Video: still images representing scenes in motion
– Still images are called frames
– Video is often accompanied by audio

• How much information is contained in video?

– Frame size (or resolution):
• How large (width x height) is each frame? Number of pixels
• How much depth in the colour? Bits per pixel
– Frame rate: how often are frames generated?
• About 15 frames per second needed to make illusion of motion

11
 Compression
• Raw data for audio/video:
– PCM voice: 8kHz sampling rate, 8 bits per sample: 64kb/s
– PCM audio (e.g. music): 44kHz, 16 bits per channel, 2 channels:
1.4Mb/s
– Standard Definition Digital TV: 720x576 pixels, 24 bits per pixel,
25 frames per second: 248Mb/s
– High Definition TV: 1920x1080 pixels: 1244Mb/s

• Compression is often used to reduce capacity needed for:

– Storage
– Transmission

12
 Compression
• Lossy compression
– Reduces the quality (lose information)
– Most commonly used today; reduce amount of data from 5%
to 25% of original size
• Lossless compression
– No loss of information, hence quality is maintained
– Compress to 50% to 70% of original size

• Amount of compression depends on algorithm and data source

– Example: high motion video cannot be compressed as
much as low motion video

13
 Codecs
• A general term referring to the software/hardware/standard that perform
modulation, compression, formatting of source into digital formats
• Trade-off in: bit rate, quality, complexity (processing time)
• Voice codecs • Audio codecs (lossless)
– ITU G.711: 64kb/s PCM – FLAC
– ITU G.722: 64kb/s ADPCM – Shorten
– ITU G.726: 16-40kb/s ADPCM – WMA Lossless
– ITU G.728: 16kb/s CELP – MPEG-4 Lossless
– ITU G.729: 8kb/s ACELP
– GSM: 14kb/s • Video codecs
– MPEG-1: VCD (1.5Mb/s)
– MPEG-2: DVD, digital TV (3-6Mb/s)
– MPEG-4: DivX, Xvid, FFmpeg
• Audio codecs (lossy)
– H.264 (MPEG-4 AVC): Bluray
– MPEG: MP3, AAC
– WMV
– Dolby Digital AC-3
– DTS • (Container formats:
– Vorbis – AVI, Ogg, WAV, MOV, MPEG4 TS/PS,
– WMA …) 14
Performance Requirements for Applications
• Data Rates
– Voice/audio applications require 10’s to 100’s kb/s
– Video applications require 100’s kb/s to 10’s Mb/s
• Errors
– Most applications can tolerate small number of errors (i.e. loss of
data)
• Can use Forward Error Correction (FEC) to reduce errors
• Re-transmission schemes are avoided because of the extra delay they
incur
– Errors result in drop in quality at receiver
• Delay
– Interactive (or conversational) applications have strict
delay requirements
• Voice call: <150ms is unnoticeable; 150-400ms is tolerable;
>400ms is unusable
– Streaming applications can tolerate large delays by using buffers
• Jitter
– Most applications require low jitter for smooth playback of
audio/video 15
 Example: Delay, No Jitter

Source sends a packet every

40ms

The network has delay of 50ms.

The jitter is 0 (every packet
experiences a delay of 50ms)

When a packet is received, the

data is played back.

There is 50ms delay between when

the source starts sending to when
the playback starts at receiver.

With no jitter, the playback is smooth.

16
Example: Delay and Jitter

There is 50ms delay between when

the source starts sending to when
the playback starts at receiver.

With jitter, the playback is no

longer smooth. There is a large
delay between the second and
third packet.
e.g. the frame in packet 2 will
be frozen until the packet 3
arrives

17
Example: Delay, Jitter and Buffering
Delay: 50±10

Send Receive Playback

0
1 The receiver stores the each packet
in a buffer (not played immediately).
50
40
2 50
The first packet is buffered for 20ms,
1 Buffer for 20
and then the frame is played back.
40 70
120 80 1 Packets are buffered when
3 2 necessary and played back at a
constant rate. Hence the playback
110
160 60 2
is smooth.
4
180 Note there is now a 70ms delay
50
3 190 between when the source starts
3 sending to when the playback starts
210
at receiver.
4
230
4
And the receiver needs extra
complexity and memory for buffering.
18
 Dealing with Jitter
• Playback buffers are the main mechanism to deal with jitter
• Packets must include sequence numbers and timestamps
• Receiver buffers packets until they are ready to be played
• How long to delay before starting playback?
– For stored audio/video streaming, several seconds in delay
buffer is ok
– For live audio/video streaming, 1-3 seconds may be tolerable
– For interactive applications, must be less than tolerated
delay (100’s of milliseconds)

19
Multimedia Applications in the
Internet

20
 The Internet offers Best-Effort Service
• IP:
– Unreliable, best-effort delivery of datagrams
– No guarantee of delivery, no timing guarantees
– No priority for different applications
– Datagrams from an application may be processed in different
ways by routers, and even take different routes
• TCP:
– Provides reliability using a retransmission scheme
• Adds considerable extra delay if errors occur
– At start of TCP connection, throughput is low (to avoid
congestion)

• The Internet, and TCP/IP, do not have built-in

mechanisms to support multimedia applications
• Yet, with additional supporting mechanisms, multimedia
applications in the Internet are successful
21 21
Voice over IP Networks

22
 Terminology
• Remember, the Internet refers to a specific IP network

• Voice over the Internet Protocol, VoIP, IP Telephony, IP phone

– Technology using IP (and other protocols) to make voice calls

• Internet telephony, Internet phone, Voice over the Internet

– Using the Internet for voice calls (using VoIP technology)

• Example:
– Skype is VoIP software; it allows voice calls over the Internet
– NTT (Japan) over a VoIP service to customers on the NTT network
– Thammasat Uni may deploy their own VoIP network
– In the case of NTT and Thammasat, they may use their private IP
network, separate from their network attached to the Internet

23
 Real-time Transport Protocol
• RTP: Transmit digitized audio/video signals over an IP network
• Uses UDP
– Consider RTP as a transport protocol
– Since TCP is not well suited to transfer of multimedia
communications, RTP was designed
• Main functionality of RTP
– Allow any type of media (voice, video using any codec) to
be transferred
– Adds a sequence number to each block of media
– Adds a timestamp to each block of media

• RTP does not provide any guarantees of reliability, timeliness

or priority mechanisms.

24
 RTP Packet Format
• Minimum size of header is 12 bytes

1 16 17 32

Ver P X CC M Payload Type Sequence Number

12 bytes

Timestamp

Synchronization Source Identifier (SSRC)

Options + Padding

Data

25
• Sequence number is used for RTP Packet Format
each packet (initial value
chosen randomly by application)
• Optional fields may be • Timestamp indicates time
included; if so, the X bit is set to when data was sampled at
1 source
• P bit is set if no padding is • SSRC is a unique ID for the
needed after payload source (multiple source
• M bit is used by applications to together transmit)
indicate if markers are • CC field indicates the number of
included (special packet mark) sources contributing to the
• Payload type indicates the stream
format of the data (payload), e.g. • Optional fields include a
– 0: PCM, 8kHz, 64kb/s Contributing Source ID
– 3: GSM, 8kHz, 13kb/s
– 14: MPEG audio, 90kHz
– 26: Motion JPEG
– 33: MPEG2 video
26
 RTP Translation and Mixing
• An application may change the payload type in the middle of a
session
-- change encoding to achieve better quality or lower data
rate
– The Payload Type field makes such translation possible
– Translation may be performed by intermediate devices
• Multiple sources may contribute to a session
– E.g. a tele-conference between group of people at one
location to a group at another location
• Each person at location A is a source; their stream of data may go to a
central mixer, which combines them together into a single stream to be
sent to the other location
– SSRC for each person would be different
• The mixer uses a new SSRC, but includes the original SSRC’s
in the optional Contributing Source ID fields and sets CC
accordingly
– Combined with mutlicast, mixing can lead to significant performance
improvements 27
 RTP Control Protocol (RTCP)
• RTP is for sending audio/video data streams
• RTCP is used for exchanging information between
senders / receivers about the streams and users
• There are 5 types of RTCP messages:
– Sender report: sender periodically sends a report to receivers;
includes at least an absolute timestamp so receivers can
synchronise different streams
– Receiver report: receiver periodically sends a report to senders;
indicate the conditions of the reception (e.g. congestion, buffer size)
allowing senders to adapt their sending rates/quality
– Source description message: sender may send information
describing the owner of the stream
– Bye message: sender sends this when ending the stream
– Application specific message: applications may use this for
their own purpose, e.g. close d captions or subtitles for a video
stream
28
 IP Telephony and Signalling
• With voice communications, signalling refers to the process
of establishing a telephone call
• In the PSTN (public service telephone network), signalling
is performed using a protocol called Signalling System 7
(SS7)
– Given a destination phone number, forms a circuit between
source and destination
– Handles call forwarding, error reporting, busy signals, …
• In IP telephony, an equivalent protocol is needed
– Must be able to translate between PSTN and IP network
using a gateway device (connecting two technology)
– Two sets of standards proposed for IP telephony signalling:
• Session Initiation Protocol (SIP) by IETF (Internet
engineering task force)
• H.323 by ITU (International telephony union)
29
 Session Initiation Protocol
• SIP provides following mechanisms for an IP network:
– Caller notifies a callee that it wants to start a call; allows
participants to agree upon codecs and end calls
– Caller determine the IP address of the callee
– Changing codecs during a call; inviting new participants to
a call; call transfer; call holding; … (change bandwidth based
on congestion)
• SIP is an Application level protocol
– Uses UDP (or TCP in special cases)
– SIP uses port 5060
– Sends text-based messages in a format similar to HTTP
– Uses addresses similar to email address, e.g.
sip:[email protected]
– Does not specify the data transfer mechanism (RTP or others
can be usedk in data transmit)
30
 SIP Example
Alice Bob
(caller) (callee)

µ Law Audio

31
 SIP Names and IP Addresses
• In IP networks, DNS maps domain names to IP addresses
• DNS works because servers are normally associated with a single
fixed IP address
• But users are often associated with multiple, dynamic IP addresses
– Static IP for PC at work; dynamic IP for PDA; dynamic IP for
PC at home
• SIP uses:
– Registrar Servers to keep track of a users current IP address
• Each user has an associated Registrar
• When a user starts their SIP client, the client informs the
Registrar Server of the IP address (if server change, notify reg:)
– Proxy Servers to handle SIP INVITE’s on behalf of users
• A caller sends an INVITE to a Proxy. The Proxy may:
– Find the IP address of the callee via the Registrar Server,
and initiate the call
– Redirect the caller to another location (e.g. voicemail or
website) 32
 SIP Example
Assumes Keith has
informed upenn.edu that
he is at eurecom.fr

INVITE [email protected]
200 OK

Data transfer

Jim (caller) Keith (callee)

33
 SIP for Voice, Video and Data
• SIP is a general protocol for initiating and managing sessions
– Voice calls
– Video calls
– Data sessions, especially instant messaging

• Most IP phones today will support SIP, RTP (real-time transmit

protocol) and RTCP(real-time transport control protocol)
– Softphone: software that implement these protocols; run on
normal computers
• Note that Skype uses its own proprietary protocol, not SIP or
RTP
– Standalone phone: hardware built for the purpose of an IP
phone
– Adapters for PSTN phones

34
Streaming Stored Audio/Video

35
 Streaming Audio/Video
• Examples:
– Internet Radio
– Web-based Video: Youtube
• Approach:
– Content (audio/video) is stored on a streaming server
• For live content, it is generated and stored on the server
– Users request content
• Clients are typically web browsers and media players
• – Media player may be standalone (e.g. Windows Media Player,
WinAmp, …) or embedded in web pages (e.g. Flash Media Player)
• Requests are either direct to streaming server or via a separate
web server
– Content is sent from streaming server to client
• Using standard (RTP, HTTP) or proprietary protocols
36
 Accessing Content on Web Server

1. Metafile describes the content: location, encoding, name, …

• Examples: ASX, RAM, PLS, SMIL, …
At first setup connection, then open media player and send metafile.
2. Web browser launches the media player and sends the metafile to player
3. Media player uses HTTP to request the content 37
 Accessing Content from Streaming Server

• Metafile accessed via web server, whereas content is accessed

via streaming server
• HTTP is not needed for content delivery
– Can use RTP or other protocols
38
Sometime file stored in webserver, sometime in
streaming server

Real-Time Streaming Protocol

• RTSP can be used for controlling the stream playback
– Start, Pause, Describe streams

IPTV

39
 Viewing TV and Videos in Networks
• Note: not limited to traditional TV programming; also includes video- on-
demand (VoD) and other content
• Three main approaches:
– Internet or Web-based Television/Video
• Using the public Internet (especially WWW) to view video
• Small image (post card sized) on PC
• Speeds less than 1Mb/s required for acceptable quality
– File Based TV/Video Distribution
• Viewed on a PC or TV
• Non-real-time (i.e. download entire file, watch at any time),
quality depends on coding
• Accessed from normal Internet, usually using P2P file sharing
– IPTV
• High quality image, real-time reception on large TV display
• Transfer requires “network in network” (much more control
than normal Internet)
– Multicasting, QoS, caching
• Separate network than Internet
40
 Broadband modem
Customers Equipment for IPTV IP set top box
Stereo speaker
TV

41
 Example IPTV Applications
• Digital Television
– Delivering existing and new digital TV content to consumers
• Video on Demand (VoD)
– Users can select specific video content, usually for a fee (similar to
“pay-per-view”)
• Business TV to Desktop
– E.g. employees view news channels or financial reporting
• Distance Learning
– Although traditional teleconference systems support lectures, IPTV
will deliver content to the individuals (rather than conference rooms)
• Corporate Communications
– Director or CEO delivering speeches to employees
• Mobile Phone TV
– With high-speed wireless data networks, the most practical way of
delivering TV to mobiles
• Video Chat
42
 IPTV using Private Networks

• Many companies are looking to deliver IPTV over

private IP networks
– Either existing IP networks for Internet access, or separate
IP networks

• Why a separate IP network?

– To deliver the quality expected for standard TV (including
high definition digital TV), require a high level of control
over network operation
– If an company (ISP, TV network, Cable company)
owns/operates the entire IP network, they can control
the performance delivered to applications

43
IPTV Network

44
 Technologies for IPTV
• Devices
– Video Headend: converts audio/video into appropriate digital
format for transmission (e.g. MPEG2, MPEG4)
– Set Top Box (STB): manage IPTV content within customers
network
• Protocols
– Video delivery: RTP, RTSP (real time streaming protocol) etc.
• Network Management and Control
– Multicast
– QoS control
– Authentication, authorisation, accounting, …
• Network Technologies
– Core Networks: SDH (synchronous digital hierarchy), optical
fibre
– Access Networks: ADSL2, optical fibre, coaxial cable, Ethernet
– Home Networks: Ethernet, wireless LAN 45
 IPTV Bandwidth Requirements
• Lets consider example scenario in a home:
– Digitized voice: 64kb/s (per voice call)
– High speed data access: 2 to 4Mb/s (per user)
– Standard Definition TV (SDTV): 2 to 4Mb/s (per channel)
• 720 x 576 (width x height) pixels
• Analog TV, Digital TV, SVCD, DVD, DV
– High Definition TV (HDTV): 8 to 10Mb/s (per channel)
• 1080 x 720, 1260 x 1080, …
• 1920 x 1080 (Full HD)
• HDTV, Blueray Discs, HD DVD
• Then a house may require 15Mb/s to 30Mb/s

• The “bottleneck” is usually the “last mile”: Service Provider

Access Network
46
 Example: Core Network Requirements

• Service Provider IP Network

47
Example: Core Requirements for Video on Demand

• With true VoD, need to use unicast (send separate stream to

individual subscribers)
• MPEG4 compression quality high. Bandwidth req low.

48
Technologies for Service Provider Access Network
• ADSL and ADSL2+
– Uses existing copper telephone lines
– Download speeds depend on distance from telephone exchange

Distance (km) ADSL (Mb/s) ADSL2+ (Mb/s)

0.3 12.5 26.0
1 12.5 25.5
2 11.0 15.5
3 7.5 7.5

– Distance Increase, bandwidth decrease

– ADSL2+ (and similar DSL technologies) are only suitable if the
termination point is close to the home (distance is short)
– Hence, fibre installations are typically need to either:
• Bring the termination point closer to the home
• Connect directly to the home (removing the need for
copper/ADSL) 49
Technologies for Service Provider Access Network

• Fibre-to-the-Node:
– Optical fibre connects to nodes or cabinets in a
neighbourhood (100’s to 1000’s of homes)
– Existing copper (ADSL) or coaxial cables (HFC) are then use
from the node to the home
• Fibre-to-the-Curb:
– Usually to the street-level, support several or 10’s of users
– Again, copper or coaxial to the home
• Fibre-to-the-Home:
– Fibre runs direct to each home (or business, building), directly
connecting to the home network
– No need for ADSL, HFC or other (much slower) alternatives

50
• Summary:
– Optical fibre can support speeds of Gb/s+
– The closer the fibre gets to home, the better
(however usually very expensive to install!)
– Other options: wireless (IEEE 802.11n), Ethernet
(especially for businesses)…

51