The Application Layer

1
The Application Layer

• Chapter 7

2
The application layer

• The layers below the application layer are there
  to provide reliable end-to-end communication.
• The application layer contains all the
  higher-level protocols.
• Supporting protocols
• DNS - Domain Name System
• Real Applications
• Email.
• World Wide Web.
• Multimedia.

3
DNS The Domain Name System

• The DNS Name Space
• Resource Records
• Name Servers

4
The DNS Name Space

• A portion of the Internet domain name space.

5
Resource Records

• The principal DNS resource records types.

6
Resource Records (2)

• A portion of a possible DNS database for cs.vu.nl.

7
Name Servers

• Part of the DNS name space showing the division
  into zones.

8
Name Servers (2)

• How a resolver looks up a remote name in eight
  steps.

9
Electronic Mail

• Architecture and Services
• The User Agent
• Message Formats
• Message Transfer
• Final Delivery

10
Electronic Mail (2)

• Some smileys. They will not be on the final exam
  -).

11
Architecture and Services

• Basic functions
• Composition
• Transfer
• Reporting
• Displaying
• Disposition

12
The User Agent

• Envelopes and messages. (a) Paper mail. (b)
  Electronic mail.

13
Reading E-mail

• An example display of the contents of a mailbox.

14
Message Formats RFC 822

• RFC 822 header fields related to message
  transport.

15
Message Formats RFC 822 (2)

• Some fields used in the RFC 822 message header.

16
MIME Multipurpose Internet Mail Extensions

• Problems with international languages
• Languages with accents (French, German).
• Languages in non-Latin alphabets (Hebrew,
  Russian).
• Languages without alphabets (Chinese, Japanese).
• Messages not containing text at all (audio or
  images).

17
MIME (2)

• RFC 822 headers added by MIME.

18
MIME (3)

• The MIME types and subtypes defined in RFC 2045.

19
MIME (4)

• A multipart message containing enriched and audio
  alternatives.

20
Message Transfer

• Transferring a message from elinore_at_abc.com to
  carolyn_at_xyz.com.

21
Final Delivery
(a) Sending and reading mail when the receiver
has a permanent Internet connection and the user
agent runs on the same machine as the message
transfer agent. (b) Reading e-mail when the
receiver has a dial-up connection to an ISP.
22
POP3

• Using POP3 to fetch three messages.

23
IMAP

• A comparison of POP3 and IMAP.

24
The World Wide Web

• Architectural Overview
• Static Web Documents
• Dynamic Web Documents
• HTTP The HyperText Transfer Protocol
• Performance Ehnancements
• The Wireless Web

25
Architectural Overview

• (a) A Web page (b) The page reached by clicking
  on Department of Animal Psychology.

26
Architectural Overview (2)

• The parts of the Web model.

27
The Client Side

• (a) A browser plug-in. (b) A helper application.

28
The Server Side

• A multithreaded Web server with a front end and
  processing modules.

29
The Server Side (2)

• A server farm.

30
The Server Side (3)

• (a) Normal request-reply message sequence.
• (b) Sequence when TCP handoff is used.

31
URLs Uniform Resource Locaters

• Some common URLs.

32
Statelessness and Cookies

• Some examples of cookies.

33
HTML HyperText Markup Language
(b)

• (a) The HTML for a sample Web page. (b) The
  formatted page.

34
HTML (2)

• A selection of common HTML tags. some can have
  additional parameters.

35
Forms

• (a) An HTML table.
• (b) A possible rendition of this table.

36
Forms (2)

• (a) The HTML for an
  order form.
• (b) The formatted page.

(b)
37
Forms (3)

• A possible response from the browser to the
  server with information filled in by the user.

38
XML and XSL

• A simple Web page in XML.

39
XML and XSL (2)

• A style sheet in XSL.

40
Dynamic Web Documents

• Steps in processing the information from an HTML
  form.

41
Dynamic Web Documents (2)

• A sample HTML page with embedded PHP.

42
Dynamic Web Documents (3)
(a) A Web page containing a form. (b) A PHP
script for handling the output of the form. (c)
Output from the PHP script when the inputs are
"Barbara" and 24 respectively.
43
Client-Side Dynamic Web Page Generation

• Use of JavaScript for processing a form.

44
Client-Side Dynamic Web Page Generation (2)

• (a) Server-side scripting with PHP.
• (b) Client-side scripting with JavaScript.

45
Client-Side Dynamic Web Page Generation (3)

• A JavaScript program for computing and printing
  factorials.

46
Client-Side Dynamic Web Page Generation (4)

• An interactive Web page that responds to mouse
  movement.

47
Client-Side Dynamic Web Page Generation (5)

• The various ways to generate and display content.

48
HTTP Methods

• The built-in HTTP request methods.

49
HTTP Methods (2)

• The status code response groups.

50
HTTP Message Headers

• Some HTTP message headers.

51
Example HTTP Usage

• The start of the output of www.ietf.org/rfc.html.

52
Caching

• Hierarchical caching with three proxies.

53
Content Delivery Networks

• (a) Original Web page. (b) Same page after
  transformation.

54
The Wireless Web

• Steps in looking up a URL when a CDN is used.

55
WAP The Wireless Application Protocol

• The WAP protocol stack.

56
WAP (2)

• The WAP architecture.

57
I-Mode

• Structure of the i-mode data network showing the
  transport protocols.

58
I-Mode (2)

• Structure of the i-mode software.

59
I-Mode (3)

• Lewis Carroll meets a 16 x 16 screen.

60
I-Mode (4)

• An example of cHTML file.

61
Second-Generation Wireless Web

• A comparison of first-generation WAP and i-mode.

62
Second-Generation Wireless Web (2)

• New features of WAP 2.0.
• Push model as well as pull model.
• Support for integrating telephony into apps.
• Multimedia messaging.
• Inclusion of 264 pictograms.
• Interface to a storage device.
• Support for plug-ins in the browser.

63
Second-Generation Wireless Web (3)

• WAP 2.0 supports two protocol stacks.

64
Second-Generation Wireless Web (4)

• The XHTML Basic modules and tags.

65
Multimedia

• Introduction to Audio
• Audio Compression
• Streaming Audio
• Internet Radio
• Voice over IP
• Introduction to Video
• Video Compression
• Video on Demand
• The MBone The Multicast Backbone

66
Multimedia

• For many people, multimedia is the holy grail of
  networking. It brings immense technical
  challenges in providing (interactive) video on
  demand to every home and equally immense profits
  out of it.
• Literally, multimedia is just two or more media.
  Generally, the term of multimedia means the
  combination of two or more continuous media. In
  practice, the two media are normally audio and
  video.

67
Introduction to Audio

• The representation, processing, storage, and
  transmission of audio signals are a major part of
  the study of multimedia systems.
• The frequency range of the human ear runs from 20
  Hz to 20K Hz. The ear is very sensitive to sound
  variations lasting only a few milliseconds. The
  eye, in contrast, does not notice changes in
  light level lasting only a few milliseconds.
  
• So, jitter of only a few milliseconds during a
  multimedia transmission affects the perceived
  sound quality more than it affects the perceived
  image quality.
• Audio waves can be converted to digital form by
  an ADC (Analog Digital Converter), as shown in
  the following figure.

68
Digital audio

• (a) A sine wave. (b) Sampling the sine wave.
  (c) Quantizing the samples to 4 bits.

If the highest frequency in a sound wave is f,
then Nyquist theorem states that it is sufficient
to make 2f samples at a frequency . Digital
samples are never exact. The error introduced by
the finite number of bits per sample is called
the quantization noise.
69
Examples of sampled sound

• Two well-known examples of sampled sound
• The telephone Pulse code modulation uses 7-bit
  (US and Japan) or 8-bit (Europe) samples 8000
  times per second, giving a date rate 56 Kbps or
  64 Kbps.
• Audio CDs are digital with a sampling rate of
  44,100 sample/sec, enough to capture frequencies
  up to 22,050 Hz. The samples are 16 bits each,
  which allows only 65,536 distinct values (but the
  dynamic range of the ear is about 1 million when
  measured in steps of the smallest audible sound.)
  Thus using 16 bits per sample introduces some
  quantization noise.
  With 44,100
  samples/sec of 16 bits each, audio CD needs a
  bandwidth of 705.6 Kbps for monaural and 1.411
  Mbps for stereo, which requires almost a full T1
  channel (1.544 Mbps) to transmit uncompressed CD
  quality stereo sound.

70
MIDI (Music Instrument Digital Interface)

• Dozens of programs exist for personal computers
  to allow users to record, display, edit, mix, and
  store sound waves from multiple sources.
• A standard, MIDI (Music Instrument Digital
  Interface), is adopted by virtually the entire
  music industry. It specifies the connector, the
  cable, and the message format.
• Each MIDI message consists of a status byte
  followed by zero or more date bytes.
• A MIDI message conveys one musically significant
  event, such as a key being pressed, slider being
  moved, or a foot pedal being released. The status
  byte indicates the event, and the data bytes give
  parameters (e.g. which key was depressed).
• The heart of every MIDI system is a synthesizer
  (often a computer) that accepts messages and
  generates music from them.
• The advantage of transmitting music using MIDI
  compared to sending a digitized waveform is the
  enormous reduction in bandwidth, often by a
  factor of 1000.

71
Audio Compression

• The most popular audio compression algorithm is
  MPEG audio, which has three layers (variants), of
  which MP3 (MPEG audio layer 3) is the most
  powerful and best known.
• Audio compression can be done in one of two
  ways
• Waveform coding the signal is transformed
  mathematically by a Fourier transform into its
  frequency components. The amplitude of each
  component is then encoded in a minimal way. The
  goal is to reproduce the waveform accurately at
  the other end in as few bits as possible.
• Perceptual coding exploits certain flaws in the
  human auditory system to encode a signal in such
  a way that it sounds the same to a human
  listener, even if it looks quite different on an
  oscilloscope. Perceptual coding is based on the
  science of psychoacoustics how people perceive
  sound. MP3 is based on perceptual coding.

72
Audio Compression

• The key property of perceptual coding is that
  some sounds can mask other sounds.
• Frequency masking the ability of a loud sound
  in one frequency bank to hide a softer sound in
  another frequency band that would have been
  audible in the absence of the loud sound.
• Temporal masking the effect that the ear turns
  down its gain when they start and it takes a
  finite time to turn it up again.

• (a) The threshold of audibility as a function of
  frequency. (b) The masking effect.

73
MP3 audio compression

• The essence of MP3 is to Fourier-transform the
  sound to get the power at each frequency and then
  transmit only the unmasked frequencies, encoding
  these in as few bits as possible.
• MP3 audio compression
• It samples the waveform at 32 kHz, 44.1 kHz, or
  48 kHz.
• The sampled audio signal is transformed from the
  time domain to the frequency domain by a fast
  Fourier transformation.
• The resulting spectrum is then divided up into 32
  frequency bands, each of which is processed
  separately.
• The MP3 audio stream is adjustable from 32 kbps
  to 448 kbps. MP3 can compress a rock'n roll CD
  down to 96 kbps with no perceptible loss in
  quality, even for rock'n roll fans. For a piano
  concert, at least 128 kbps are needed.

74
Streaming Audio

• A straightforward way to implement clickable
  music on a Web page.

75
Streaming Audio (2)
When packets carry alternate samples, the loss of
a packet reduces the temporal resolution rather
than creating a gap in time.
76
Streaming Audio (3)

• The media player buffers input from the media
  server and plays from the buffer rather than
  directly from the network.

77
Streaming Audio (4)

• RTSP commands from the player to the server.

78
Internet Radio

• A student radio station.

79
Voice over IP

• The H323 architectural model for Internet
  telephony.

80
Voice over IP (2)

• The H323 protocol stack.

81
Voice over IP (3)

• Logical channels between the caller and callee
  during a call.

82
SIP The Session Initiation Protocol

• The SIP methods defined in the core specification.

83
SIP (2)

• Use a proxy and redirection servers with SIP.

84
Comparison of H.323 and SIP
85
Video

• The human eye has the property that when an image
  is flashed on the retina, it is retained for a
  few milliseconds before decaying.
• If a sequence of images is flashed at 50 or more
  images/sec, the eye does not notice that it is
  looking at discrete images. All TV systems
  exploit this property to produce moving pictures.
  
• To represent the two-dimensional image as a
  one-dimensional voltage as a function of time,
  the scanning pattern used by both the camera and
  the receiver is shown in the following figure.

86
Video Analog Systems

• The exact scanning parameters very from country
  to country
• US and Japan 525 scan lines (only 483 lines
  displayed), a horizontal to vertical aspect ratio
  of 43, and 30 frames/sec.
• Europe 626 scan lines (only 576 lines
  displayed), the same aspect ratio of 43, and 25
  frame/sec.

87
Interlacing and progressive techniques

• The interlacing technique Instead of displaying
  the scan lines in order, first all the odd scan
  lines are displayed, then the even ones are
  displayed. Each of these half frames is called a
  field. Experiments have shown that although
  people notice flicker at 25 frames/sec, they do
  not notice it at 50 fields/sec.
• Non-interlaced TV or video is said to be
  progressive.

88
Color video

• Color video uses three beams moving in unison,
  with one beam for each of the three additive
  primary colors red, green, and blue (RGB). For
  transmission on a single channel, the three color
  signals must be combined into a single composite
  signal.
• To ensure that programs transmitted in color be
  receivable on existing black-white TV sets, the
  simplest scheme, just encoding the RGB signals
  separately, was not acceptable, leading to
  incompatible color TV systems in different
  countries
• NTSC standard by the US National Television
  Standards Committee,
• SECAM (Sequential Couleur Avec Memoire) used in
  France and Eastern Europe, and
• PAL (Phase Alternating Line) used in the rest of
  Europe.
• All three systems linearly combine the RGB
  signals into a luminance (brightness) signal, and
  two chrominance (color) signals.
• Since the eye is much more sensitive to the
  luminance signal than to the chrominance signals,
  the latter need not be transmitted as accurately
  and they can be broadcast in narrow bands at
  higher frequencies.
• HDTV (High Definition TeleVision) produces
  sharper images by roughly doubling the number of
  scan lines and has an aspect ratio of 169
  instead of 43 to match them better to the format
  used for movies (using 35 mm film).

89
Digital video systems

• The representation of digital video consists of
  
• A sequence of frames.
• Each frame consists of a rectangular grid of
  picture elements, or pixels.
• Each pixel can be presented by a single bit
  (b/w), 8 bits (for high-quality b/w video), or 24
  bits (8 bits for each of the RGB colors).
• Smoothness of motion is determined by the number
  of different images per second, whereas flicker
  is determined by the number of times the screen
  is painted per second.
• To produce smooth motion, digital video, like
  analog video, must display at least 25
  frames/sec.
• Since good quality computer monitors often rescan
  the screen from images stored in memory at 75
  times/sec or more, interlacing is not needed.
  Just repainting the same frame three times in a
  row is enough to eliminate flicker.

90
Digital systems

• Current computer monitors all use the 43 aspect
  ratio so they can use inexpensive, mass-produced
  picture tubes for the TV market. Common
  configurations are
• 640 ? 480 (VGA)
• 800 ? 600 (SVGA)
• 1024 ? 768 (XGA)
• An XGA display with 24 bits per pixel and 25
  frames/sec needs to be fed at 472 Mbps, which is
  larger than the bandwidth of an OC-9 SONET
  carrrier.
• Digital video transmits 25 frames/sec but have
  the computer to store each frame and paint it
  twice to eliminate flicker.
• Analog TV broadcast transmits 50 fields(25
  frames)/sec but uses interlacing to eliminate
  flicker because TV sets do not have memory and
  cannot convert analog frames into digital form.

91
Data compression

• Transmitting multimedia material in uncompressed
  form is completely out of the question.
• Fortunately, a large body of research over the
  past few decades has led to many compression
  techniques and algorithms that make multimedia
  transmission feasible.
• All compression systems require two algorithms
• one for compressing the data at the source
  (encoding), and
• another for decompressing it at the destination
  (decoding).

92
Asymmetries of data compression algorithms

• These algorithms have certain asymmetries
• For many applications, e.g., a multimedia
  document, a movie will only be encoded once (when
  it is stored on the multimedia server) but will
  be decoded thousands of times (when it is viewed
  by customers).
  Consequently, the decoding algorithms should be
  very fast and simple, even at the price of making
  encoding slow and complicated.
  On the
  other hand, for real-time multimedia, such as
  video conferencing, slow encoding is
  unacceptable, so different algorithms or
  parameters are used.
• The encode/decode process need not be invertible
  for multimedia documents, unlike the compressed
  file transfer which ensures the receiver to get
  the original file back. When the
  decoded output is not exactly equal to the
  original input, the system is said to be lossy,
  otherwise it is lossless. Lossy
  systems are important because accepting a small
  amount of information loss can give a huge payoff
  in terms of the compression ratio possible.

93
Entropy encoding

• Compression schemes can be divided into two
  general categories entropy encoding and source
  encoding.
• Entropy encoding just manipulates bit streams
  without regard to what the bits mean. It is a
  general, lossless, fully reversible technique,
  applicable to all data.
• Run-length encoding
• In many kinds of data, strings of repeated
  symbols (bits, numbers, etc) are common. These
  can be replaced by
• a special marker not otherwise allowed in the
  data, followed by
• the symbol comprising the run, followed by
• how many times it occurred.
• If the special marker itself occurs in the data,
  it is duplicated (as in character stuffing).

94
An example of entropy encoding

• A sequence of bits
• 31500000000000084587111111111111163546740000000000
  00000000000065
• If A is used as the marker and two-digit numbers
  are used for the repetition count, the above
  digit string can be encoded as
• 315A01284587A1136354674A02265
• A saving of about 50.
• In audio, silence is often represented by runs of
  zeros. In video, runs of the same color occur in
  shots of the sky, walls, and many flat surfaces.
  All of these runs can be greatly compressed.

95
Statistical and CLUT encoding

• Statistical encoding
• Basic idea using a short code to represent
  common symbols and long ones to represent
  infrequent ones.
• Huffman coding and Ziv-Lempel algorithm used by
  the UNIX Compress program use statistical
  encoding.
• CLUT (Color Look Up Table) encoding
• Suppose that only 256 different color values are
  actually used in the image (e.g., a cartoon or
  computer-generated drawing). A factor of almost
  three compression can be achieved by
• building a 768-byte table listing the three RGB
  values of the 256 colors actually used, and then
• representing each pixel by the index of its RGB
  value in the table.
• This is a clear example where encoding (searching
  the whole table) is slower than decoding (a
  single indexing).

96
Source encoding

• It takes advantages of properties of the data to
  produce more (usually lossy) compression.
• Example 1 differential encoding
• A sequence of values (e.g., audio samples) are
  replaced by representing each one as the
  difference from the previous value.
• Example 2 transformations
• By transforming signals from one domain to
  another, compression may become much easier.
  E.g., the Fourier transformation.

97
The JPEG Standard
JPEG (International Standard 10918) was developed
by photographic experts for compressing
continuous-tone still pictures (e.g.,
photographs).

• The operation of JPEG in lossy sequential mode.

98
Step 1 Block preparation

• For a given RGB input arrays shown in (a),
  separate matrices are constructed for the
  luminance, Y, and the two chrominance, and (for
  NTSC), according to the following formulas
• Y 0.30R 0.59G 0.11B
  I 0.60R - 0.28G - 0.32B Q 0.21R - 0.52G
  0.31B
• Square blocks of four pixels are averaged in the
  I and Q matrices to reduce them to 320 ? 240.
• 128 is subtracted from each element of all three
  matrices.
• Each matrix is divided up into 8 ? 8 blocks. The
  Y matrix has 4800 blocks the other two have 1200
  blocks each, as shown in (b).

99
Step 2 DCT (Discrete Cosine Transform)

• Apply DCT to each of the 7200 blocks separately.
  The output of each DCT is an matrix of DCT
  coefficients.
• DCT element (0, 0) is the average of the block.
  Other elements tell how much spectral power is
  present at each spatial frequency. These elements
  decay rapidly with distance from the origin, (0,
  0), as suggested in the figure.

• (a) One block of the Y matrix.
  (b) The DTC coefficients.

100
Step 3- 4 Quantization and differential
quantization
Quantization wipes out the less important DCT
coefficients by dividing each of the elements in
the matrix by a weight taken from a table, as
illustrated in Figure below. 
Differential quantization reduces the (0, 0)
value of each block by replacing it with the
amount it differs from the corresponding element
in the previous block. Since these elements are
the averages of their blocks they should change
slowly, so taking the differential values should
reduce most of them to small values. No
differentials are computed for other elements.
101
Step 5 - 6 run-length encoding and Huffman
encoding
The output matrix of the differential
quantization is scanned in a zig zag pattern, and
the run-length encoding is used to reduce the
output string of numbers.

• Huffman encoding is used to encode the numbers
  for storage or transmission.
• Decoding a JPEG image requires running the
  algorithm backward.
• Since JPEG often produces a 201 compression or
  better, it is widely used.

102
The MPEG (Motion Picture Experts Group) Standard
MPEG is the main algorithm used to compress
videos and has been international standard since
1993. The following discussion will focus on MPEG
video compression. MPEG-1 (International
Standard 11172) has the goal to produce video
recorder-quality output ( for NTSC) using a bit
rate of 1.2 Mbps. MPEG-1 can be transmitted over
twisted pairs for modest distances (100 meters).
MPEG-2 (International Standard 13818) was
originally designed for compressing broadcast
quality video into 4 to 6 Mbps, so it could fit
in a NTSC or PAL broadcast channel. Later, it was
extended to support HDTV. It forms the basis for
DVD and digital satellite TV. MPEG-4 is
for medium-resolution video-conferencing with low
frame rates (10 frames/sec) and at low bandwidths
(64 kbps).
103
The MPEG-1system
MPEG-1 has three parts audio, video, and system,
which integrates the other two, as shown below.

• The system clock runs at 90-kHz and feeds
  timestamp to both encoders. The timestamps are
  represented in 33 bits, to allow film to run for
  24 hours without wrapping around. These
  timestamps are included in the encoded output and
  propagated to the receiver.

104
The MPEG-1 video compression

• It exploits two kinds of redundancies in movies
  spatial and temporal.
• Spatial redundancy can be utilized by simply
  coding each frame separately with JPEG. In this
  mode, a compressed bandwidth in the 8- to 10-Mbps
  range is achievable.
• Additional compression can be achieved by taking
  advantage of the fact that consecutive frames are
  often almost identical.
• MPEG-1 output consists of four kinds of frames
• 1. I (Intracoded) frames Self-contained
  JPEG-encoded still pictures.
• It is needed for three reasons to enable start
  viewing in the middle, to enable decoding in face
  of errors in previous frames, and to enable a
  fast forward or rewind.

105
The P (Predictive) frames
2. P (Predictive) frames Block-by-block
difference with the last frame. It is based on
the idea of macroblocks, which cover pixels in
luminance space and pixels in chrominance space.
An example of where P-frames would be useful is
given in the following figure.

• A macroblock is encoded by searching the previous
  frames for it or something only slightly
  different from it.
• If a macroblock is found, it is encoded by taking
  the difference with its value in the previous
  frame, and then applying JPEG onto the
  difference.
• If a macroblock is not found, it is encoded
  directly with JPEG, just as an I-frame.
• The value for the macroblock in the output stream
  is then
• the motion vector (how far the macroblock moved
  from its previous position in each direction),
  followed by
• the JPEG encoding.

106
B (Bidirectional) and D (DC-coded) frames
3. B (Bidirectional) frames Differences with the
last and next frame. Similar to P-frames, except
that they allow the reference macroblock to be in
either a previous frame or in a succeeding frame.
B-frames give the best compression, but not all
implementations support them. 4. D (DC-coded)
frames Block average used for fast forward.
Each D-frame entry is just the average value of
one block, with no further encoding, making it
easy to display in real time. D-frames are only
used to make it possible to display a
low-resolution image when doing a rewind or fast
forward.
107
MPEG-2

• MPEG-2 differs from MPEG-1 in the following
  aspects
• It supports I-frames, P-frames, and B-frames, but
  not D-frames.
• The discrete cosine transformation uses a 10 10
  block instead of 8 8 block, to give 50 percent
  more coefficients, hence better quality.
• MPEG-2 is targeted at broadcast TV as well as DVD
  applications, so it supports both progressive and
  interlaced image, whereas MPEG-1 supports only
  progressive image.
• MPEG-2 supports four resolution levels
• low 352 ? 240 for VCRs and backward compatible
  with MPEG-1.
• main 720 ? 480 for NTSC broadcasting.
• high-14401440 ? 1152 for HDTV
• high 1920 ? 1080 for HDTV
• For high quality output, MPEG-2 usually runs at 4
  8 Mbps.

108
Video on Demand

• Watching video movies on demand is but one of a
  vast array of potential new services possible
  once wideband networking is available.
• Two different models of video on demand
• Video rental store model Users are allowed to
  start, stop, and rewind videos of theirr choice
  at will. In this model, the video provider has
  to transmit a separate copy to each one.
• Multi-channel (e.g., 5000-channel) cable TV
  model Users are not allowed to pause/resume a
  video, but they can simply switch to another
  channel shown the same video but 10 minutes
  behind. In this model, the video provider can
  start each popular video, say, every minutes, and
  run these nonstop. This model is called near
  video on demand.
• The general system structure of video-on-demand
  is illustrated in the following figure.

109
Overview of a video-on-demand system
How these pieces will fit together and who will
own what are matters of vigorous debate within
the industry. Below we will examine the design of
the main pieces the video servers and the
distribution network.
110
Video Servers
Storage capacity requirement The total number of
movies ever made is estimated at 65,000. When
compressed in MPEG-2, a normal movie occupies
roughly 4 GB of storage, so 65,000 of them would
require about 260 terabytes (without counting all
the old TV programs ever made, sports films,
etc.). Equipment and price A DAT magnetic tap
can store 8 GB (two movies) at a cost of about 5
dollars/gigabyte (the cheapest). Large
mechanical tape servers that can hold thousands
of tapes and have a robot arm for fetching any
tape and insert it into a tape drive are
commercially available now. The problem with
these systems is the access time, the transfer
rate, and the limited number of tape drives.
111
Video server storage hierarchy
Fortunately, experience with video rental stores,
public libraries, and other such organizations
shows that not all items are equally popular.
Zipf's law the most popular movies is times as
popular as the number movie. The fact that some
movies are much more popular than others suggests
a possible solution in the form of a storage
hierarchy, as shown below.
112
DVD, Disk and RAM
An alternative to tape is optical storage.
Current DVDs hold only 4.7 GB, good for one
movies, but next generation will hold two MPEG-2
movies. DVD seek times (50 msec) are slow
compared to magnetic disks (5 msec), but their
advantages are low cost and high reliability.
Suitable for holding more heavily used movies.
Disks have short access time (5 msec), high
transfer rates (320 MB/sec for SCSI 320), and
substantial capacities (gt 100 GB), which makes
them well suited to holding movies that are
actually being transmitted. Their main drawback
is the high cost for storing movies that are
rarely accessed. RAM is the fastest storage
medium but the most expensive. It is best suited
to movies for which different parts are being
sent to different destinations at the same time
(e.g., true video-on-demand with 100 users who
all started at different times). When RAM prices
drop to 50/GB, a 4-GB movie occupy 200 dollars
worth of RAM, so having 100 movies in RAM will
cost 20,000 dollars for the 200 GB of memory.
Keeping all watched movies in RAM begins to look
not only feasible, but a good design.
113
The hardware architecture of a typical video
server
114
Video server software

• The CPUs are used for accepting user requests,
  locating movies, moving data between devices,
  customer billing, etc. Many of them are time
  critical, so a real-time operating system is
  needed for CPUs.
• A real-time system normally breaks work up into
  small tasks, each with a known deadline. The
  scheduler can run an algorithm such as nearest
  deadline next.
• The interface between the video server and
  clients (i.e., spooling servers and set-top
  boxes). Two popular designs
• The traditional file system (such as UNIX) model
  the clients can open, read, write, and close
  files.
• The video recorder model the clients can open,
  play, pause, fast forward, and rewind files.
• The disk management software takes charge of
• placing movies on the magnetic disk when they
  have to be pulled up from optical or tape
  storage, and
• handling disk requests for the many output
  streams.
• Two possible ways of organizing disk storage
• Disk farm each drive holds a few entire movies.
  For performance and reliability reasons, each
  movie may be present on more than one drive.
• Disk array or RAID (Redundant Array of
  Inexpensive Disks) each movie is spread out over
  multiple drives block 0 on drive 0, ..., n 1
  block on drive n - 1, then block n on drive 0,
  and so forth. This organization is called
  striping.

115
The distribution network
The distribution network is the set of switches
and lines between the source and destination.
The main requirements imposed on the backbone
are high bandwidth. Low jitter used to be a
requirement as well, but with even the smallest
PC now able to buffer 10 sec of high-quality
MPEG-2 video, low jitter is not a requirement
anymore. Local distribution is highly chaotic,
with different companies (telephone, cable TV,
etc) trying out different networks in different
regions. ADSL (Asymmetric Digital
Subscriber Line) was the telephone industry's
first entrant in the local distribution
sweepstakes, which make use of existing copper
twisted pairs, as discussed in Chap 2. It is not
fast enough (4 8 Mbps) except for very short
local loops. Another design is to run fiber into
everyone's house, called FTTH (Fiber To The
Home), which is very expensive and will not
happen for years. When it really happens, itd be
possible for every family member to run his or
her own personal TV station!
116
FTTC (Fiber To The Curb)
About 16 copper local loops can terminate in an
end office. Able to support MPEG-1 and MPEG-2
movies. Video-conferencing for home workers and
small business is now possible because FTTC is
symmetric
117
HFC (Hybrid Fiber Coax)
Instead of using point-to-point local
distribution networks, a completely different
approach is HFC (Hybrid Fiber Coax), which is
preferred solution currently being installed by
cable TV providers, as illustrated be low.
118
HFC (Hybrid Fiber Coax)

• The current 300- to 450-MHz coax cables will be
  replaced by 750-MHz coax cables, upgrading the
  capacity from 50 to 75 6-MHz channels to 125
  6-MHz channels.
• 75 of the 125 channels will be used for
  transmitting analog TV. The 50 new channels will
  each be modulated using QAM-256, which provides
  about 40 Mbps per channel, giving a total of 2
  Gbps of new bandwidth.
• Each cable runs past about 500 houses, and each
  house can be allocated a dedicated 4 Mbps
  channel, which can be used for some combination
  of MPEG-1 programs, MPEG-2 programs, upstream
  data, analog and digital telephony, etc.
• HFC uses a shared medium without switching and
  routing. As a result, HFC providers want video
  servers to send out encrypted streams, so
  customers who have not paid for a movie cannot
  see it.
• One the other hand, FTTC is fully switched and
  does not need encryption because it adds
  complexity, lowers performance, and provides no
  additional security in their system.
• For all these local distribution networks, it is
  likely that each neighborhood will be outfitted
  with one or more spooling servers, which may be
  preloaded with movies either dynamically or by
  reservation.

119
The MBone The Multicast Backbone
MBone can be thought of as Internet radio and TV.
Its emphasis is on broadcasting live audio and
video in digital form all over the world via the
Internet. MBone has been operational since early
1992, and has been used for broadcasting many
scientific conferences, such as IETF meetings, as
well as newsworthy scientific events, such as
space shuttle launches. Most of the research
concerning MBone has been about how to do
multicasting efficiently over the
(datagram-oriented) Internet. Little has been
done on audio or video encoding. Technically,
MBone is a virtual overlay network on top of the
Internet, as shown below.
120
Major MBone components

• Mrouters (mostly just UNIX stations running
  special user-level software) are logically
  connected by tunnels (defined just by tables in
  the mrouters).
• MBone packets are encapsulated within IP packets
  and sent as regular unicast packets to the
  destination mrouter's IP address.
• Tunnels are configured manually. For a new island
  to join MBone,
• the administrator sends a message announcing its
  existence to the MBone mailing list, and
• the administrators of nearby sites then contact
  him to arrange to set up tunnels.
• Multicast addressing
• To multicast an audio or video program, a source
  must first acquire a class D multicast address
  (from a distributed database), which acts like a
  station frequency or channel number.
• Multicast group management
• Periodically, each mrouter sends out broadcast
  packet limited to its island asking who is
  interested in which channel.
• Hosts wishing to receive one or more channels
  send another packet back in response.
• Each mrouter keeps a table of which channels it
  must put out onto its LAN.

121
Major MBone components

• Multicast routing
• When an audio or video source generates a new
  packet, it multicasts it to its local island
  using the hardware multicast facility.
• This packet is picked up by the local mrouter,
  which copies it into all the tunnels to which it
  is connected.
• Each mrouter getting such a packet checks the
  routing table (based on the distance vector
  routing algorithm) and uses the reverse path
  forwarding algorithm to decide whether to drop
  and forward it.
• Moreover, the IP Time to live field is also used
  to limit the scope of multicasting. Each tunnel
  is assigned a weight. A packet is only passed
  through a tunnel if its has enough weight.
• Much research has been devoted to improving the
  above MBone routing algorithm. Read the text and
  references for more details.
• All in all, multimedia is an exciting and rapidly
  moving field. New technologies and applications
  are announced daily, but the area as a whole is
  likely to remain important for decades to come.