This chapter does not pretend to make an exhaustive study about data networks, but offer an
outline on the bases and protocols that affect the processing and transfer of video.
1.1 GENERAL CONCEPTS
A communication network can be defined as a nodes and links set between two or more points
that provide a service. A communication service it can be understood as an information
functions set that an organization provide to a user set providing them telecommunications.
A communication network must provide a completely transparent service to users, when a bit
is sent from a emitter station to a destination, the time needed to the transfer this bit has to
approach zero and it always be the same time without relying on when was sent. In an ideal
system, transfer time is zero, and the differences between transmission times at different
moments are also zero, but in reality the ideal network does not exists so is needed to refine
and define this times.
The service provided by a network will be better when a greater transparency offered by the
network to users, when less looks that this network exists. The features that define the service
quality of a network are (among others):
Bandwidth. Defined as the traffic that the network can manage in a time unit. It
gives an idea of the transfer speed between terminals. When data is introduced
on a network, headers are added to the user’s data, so it can be defined as
effective network speed as the user traffic that the network can manage, this
speed always will be less than the bandwidth of the network.
Delay. There are many ways to define this time, is usually defined as the time
difference of arrival of the first bit of a packet and its departure from the emitter.
Another measure is the time difference between a bit arrives and the output of the first
Jitter. Difference in delay times
Error Rate. Number of bit errors or erroneous packets by the amount of information
This set of data is what is called Quality of Service (QoS) that the network provides to the
users. This quality of service is easy to implement in corporative networks or in the providers
of the network service through service agreements between those suppliers and customers,
but it is hard to deploy in the internet where it performs what is known as best effort to
provide a service.
Depending on the treatment that the networks made to route data from one extreme to
another one, networks can be classified as connective and not connective.
Dependiendo del tratamiento que realicen las redes para encaminar los datos de un extremo a
otro, se pueden clasificar las redes como conectivas y no conectivas.
Connective networks establish a communication path between two extremes and all traffic
between the terminals follow this path, in the not connective network data packets can take
different paths each one.
On connective networks, the connection is made through a circuit establishment packet which
specifies the source address and destination address. The network nodes query the routing
table and create an entry on the switching table, so that in the switching tables on
intermediate nodes is creating a virtual path for communication. A connection identifier is
given to the terminals, since then, the data that is sent from one terminal to another, carries
the connection identifier that is used by nodes to switch data packets, so the commutation
table is used but the routing table not.
The not connective networks just have routing table, they do not use switching tables since
switching paths are not set for connections. The terminals sends data into packets which
specifies the source and destination address, the intermediate nodes can send a packet by one
link or another depending on different routing factors, so that packets may go by different
The comparison between these two types of networks is:
On connective networks the connection time is higher than not connective
ones, but once the connection is made, the traffic is faster than the not
connectives networks, because they do not have to query routing tables, just
the switching tables which tend to be smaller since they only have the currently
On connective networks, the packets arrives in a ordered way, while in the not
connectives ones can arrive unordered because for the network, each packet is
independent from the previous one and may be routed by different paths.
The application scope of the network in terms of extents of links between its nodes, can
determine the technology needed for each network. Depending on the of the networks
extension can be classified as:
WAN (Wide Area Network). Aimed at managing traffic between cities and
MAN (Metropolitan Area Network). Manage traffic within a city.
LAN (local Area Network). Meet the needs of a user group close together.
PAN (Personal Area Network). Are the kind of network with small communication
radio, are used to communicate personal devices.
For each scope there are different technologies and regulations and although there are
technologies in various fields, there are different specifications depending on how will be use
A W placed before the Acronyms, means that they are wireless networks, so WPAN refers to
Personal Wireless Networks, what is really apply on these networks.
1.2 THE LAYER MODEL
The OSI model of communication protocols is composed of seven levels, each one receive a
request from the upper layer, the layer has to encapsulate the data which it has to transmit
and send the data to the lower layer. In the remote terminal, data is collected by the lower
level and delivers them to higher levels. In the Illustration 4.1 it can be see the OSI protocols
stack, the terminals can have the seven levels commented above and the communication
networks support three levels: physical, link and network layer.
Illustration 4-1 OSI LEVELS
Each level behave on a way that hidden the features of the lower level, then each level is a
upper abstraction level free of the lower levels features.
In the case of TCP/IP architecture, the contemplated levels are shown in the illustration 4-2. It
can be seen that the communication network component still has the same three levels, but in
the terminal's components, the session, presentation and application are considered an
unique level called as application level.
Illustration 4-2 INTERNET LEVELS
The communication protocols play a key role in communications; these protocols are
standardized by different standardization organizations. Each level has a set of protocols
established to do on one or another form his tasks.
1.2.2 LEVEL 1 PHYSICAL
It is the lowest level on Communications architectures. It has the ultimate responsibility of
transmitting the bits on a given physical medium. So the specifications will have to define:
How the electrics signals will be modulated to transforms the bits to electric,
magnetic, or optical levels.
Mechanic, electric, electromagnetic specifications.
Basic errors detection.
Network’s physical architecture: point to point lines, multipoint, etc.
Parallel or serial transfer.
Examples: Coaxial cables, twisted pair, RS422, RS232, microwave, fiber, radio…
Communications equipments at this level are in charge to physically connect the terminals, like
the hubs, or to regenerate the signal when it reach his distance limit (repeaters).
1.2.3 LEVEL 2 LINK
This level is in charge of data government between communications nodes. It is divided in two
Media Access Control. Is the closest part to physical level, is responsible for
building the frame that is sent to the physical level, adding to the frame, the
link MAC address.
Link Control. Pick up the packet from network level and prepare the frames to the
Media Access Control sublevel.
This level is responsible for managing the necessary operations for the management of the
link, about each node how must to behave. It can be in master-slave or peer-to-peer mode.
This level defines the kind of message that is send from one terminal to another. It can work
with character oriented protocols in which controls are made by control characters (SYN, ETX,
DEL, etc.) or by protocols block oriented.
It performs error detection on a greater level than the physical level.
As protocols examples in this level are: HDLC,LLC, PPP, LAPD, LAPm, LAPX, etc.
MAC 802.3 802.11 802.15 802.15.4
PHYSICAL L. Ethernet Wi-Fi Wimax Zigbee
Illustration 4-3. PHISICAL LAYER
In the illustration 4-3 it can observe how different network recommendations define the
physical layer and also the Media Access Control sub layer at link level. They share the access
control layer, so it can be connected networks with Ethernet, WiMax or Zigbee technology
without connecting routers between those networks.
The devices working at this level have the intelligence to switch packets at the MAC address
level, these devices are the bridges and switches.
1.2.4 LEVEL 3 NETWORK
This level provides the connectivity between the network’s endpoints, the computers
connected to this network. The primary mission of this layer is to take data from one point to
another. The telecommunications companies only provide services from layer 1 to 3; from this
layer the protocols are only in computers.
For routing, networks are divided into two types depending on the way to route packets:
Connection oriented: virtual circuits. The first message arrives to destination
address; the network creates a virtual circuit between devices and sends the
virtual circuit identifier to emitter. From this moment, the emitter sends the
content with the virtual circuit identifier.
Not connection oriented: Datagrams. All messages coming from the emitter, carries
the destination address, so the nodes, all times, has to decide where they should send the
Protocols in this level are: IPv4, IPv6, IPX, AppleTall, CSL, DECNet. The best known are the IP
family that provide not connection oriented connectivity. That means that it does not take
responsibility about packets to arrive at their destination or arrive in order. In each node is
necessary to calculate the route.
It is the most extended protocol. The illustration 4-4 shows an IP datagram header. To data
sent by level 4 is added a header to transmit the packet on the network.
Priority D T R Unused
Ver HLen Tos Total Lenght
Identification Flags Fragment desp.
TTL Protocol Checksum
Source IP Address
Destination IP Address
IP options Filler
Level 4 data
Illustration 4-4 IP HEADER
1.2.5 LEVEL 4 TRANSPORT
It enables applications to communicate to each other, the protocols at this level:
TCP (Transmission Control Protocol).‐ Connection oriented, so that guarantees
delivery and order packets.
UDP (User Datagram Protocol).‐ is a not connection oriented protocol, so it does not
guarantees delivery and order packets.
1.3 VIDEO TRANSFER ORIENTED PROTOCOLS
For computer networks, the Ethernet technology is gaining to the others networks
technologies because of its flexibility, scalability and Price. At the network level, the most
widespread protocol is IP protocol. For video transmission, it is also producing this migration.
At the network level, as they are making these changes, it is using the IP protocol; at link and
physical level, depends on the type of transmission on which the transfer is in progress.
1.3.1 PHYSICAL AND LINK LEVELS
One of the networks that define these two levels is the Ethernet network. This network is most
commonly used for connectivity between computers, not only in LAN environments but also is
going to the local loop, MAN and WAN environments, removing leadership to other traditional
networks for this type of environment.
The original idea came from the Aloha network from Hawaii in which radio stations broadcast
in the same channel when they need to transmit. On this basis, in 1973 Robert Metcalfe
devised variations on this simple basis for implementation this protocol on networks
environments over a physical medium and added the possibility that the transmitter is
listening to the media and transmits only when it is free, the same transmitter could be heard
and if it detects collisions because another device is also transmitting, waits and transmits
again when watch the channel free after a certain time.
This simple philosophy clashed with the complexity of protocols and network connections that
existed at that time, and from a single medium concept was evolved to the configuration that
is used today by point to point connections between computers and switches, configuring a
star network, allowing collision-free transmission, a staple for video transmission in real time.
IEEE specifies recommendations for Ethernet over the 802.3X standards series. These
standards define the links from 10Mbps up to 100Gbps and cable or fiber environments.
1.3.2 STREAMING AND FILE COPY.
Transferring images between remote computers, is done by file transfer or streaming.
File transfer means a reliable copy of the file from source to destination; the video is stored in
the destination and the packets that arrive wrong in the transmission channel are sent again.
ILLUSTRATION 4-5 FILE COPY AND STREAMING TRANSFER
Streaming video transfer means that the video is transmitted in real time. This restriction
requires two actions that degrade the image signal. First, the video will be coded the same
speed that the line, so that the image quality is will be limited by that speed. Second, the
packets that are wrong in the transmission line are not sent again and these errors will be
hiding with algorithms that attempt to reconstruct the original signal based on image
correlation. Is the price to pay for real-time transfer, adjust the quality of the line speed and
degradation of the image by transmission errors.
In a file transfer, the file it can be send at original quality, without having to transcode at lower
speeds and for errors in transmission, is solved with an increase in transfer time.
In some settings that work with large files and with a tight transfer time, the file transfer time
can be unpredictable and requires to transcoding the material and send it in streaming to
ensure that it is delivered in a controlled time.
The illustration 4-5 shows the difference between streaming and file copying, while streaming
images are displayed real-time as they data come, the file copying requires a full copy of the
file for later playback.
1.3.3 FTP (FILE TRANSFER PROTOCOL)
Is the file transfer protocol more widespread, but despite that is set
as a protocol designed to transfer video, is a generic protocol for the
transfer and is widely used in video as in other environments.
The protocol is based on connection-oriented service provided by the
Illustration 4-6 FTP TCP from level 4 as shown in illustration 4.6.
STACK Based on client-server concept, there are applications on both the
server and client of two types of transfers PI (Protocol Interpreter)
and DTP (Transfer Process), see illustration 4.7.
ILLUSTRATION 4-7 FTP SERVICES APPLICATIONS
PI clients commands specify the port to be used in the transfer, the form of data transfer and
the operation performed (Retrieve, List, Store, etc.). The types of commands are:
Access control (USER, PASS, ACCT, CWD,CDUP, SMNT, QUIT)
Transfer settings (PORT, PASV, TYPE, STRU, MODE)
From FTP service (RETR, STOR, STOU, DELE, LIST, …………….).
The transfer of data is made between the user and DTP server.
There are applications that implement a user interface for automate transactions between the
client and PI service, in illustration 4-7 presents a snapshot of a freely distributed application
that manages and automates the transfer of commands and data on an ftp session.
ILLUSTRATION Error! No text of specified style in document.4-8 FTP APPLICATION
1.3.4 FTP ALTERNATIVES
The main problem underlying the ftp transfer is the communication channel in the lower
levels. As seen in previous sections, the networks that are being imposed are data transfer
based on not connection oriented networks (IP) on which implements a connection-oriented
at own computer with TCP. Since different levels hide information from lower levels to higher,
TCP has no information on the type of links on the lower layers so it does not know the
network delays and sliding window sizes in use can be valid for one type of links, but not for
To solve the problem of transfer time, there are three types of methods or variations of the
UDP based transfers.
Parallel TCP sessions.
Ạbout the first group, which is based on UDP instead of TCP, there are several commercial
applications based on patents using the UDP protocol for transferring packets numbered, this
transfer is done continuously, without taking into account the transmission delay or windows
so that the application is responsible for requesting retransmissions deemed necessary to
arrive content at the destination and to present the data in order. These tools transfer
management system, planning, monitoring, encryption, prioritization, transfers, one by one,
one to many, etc. The most popular applications are based on patents: Aspera, Signiant,
FileCatalyst, RocketStream, DigitalRapids, Kencast. There are also free distribution: UDT, UFTP,
In the second group are the programs that try to adjust the size of the transmission window to
the delay before the transfer is sent to the destination table and measure the time it takes to
return the message, so you get the time delay and the ideal transmission window. In this group
fall TCP-HighSpeed, Scalable, BIC, FAST, H-TCP and L-TCP.
In parallel TCP sessions, the file to send is crumbling in many ftp sessions and each ftp has to
make a few transfers. Examples of this group are PSockets and GridFTP.
One system often used is the use of dedicated hardware, which compresses the data to
transfer smaller files. In the case of image files, such systems do not give good result because it
is hard to achieve further compress in video files.
As an example of one of these protocols we will see the UFTP, which is a protocol that uses the
UDP channel for sending files. It is currently used by the Wall Street Journal for the distribution
of its sites by satellite. The protocol is designed to send large volumes of information to
multiple receivers and works in three phases:
1. Transfer announcement
2. Transfer phase and NACK management
3. Confirmation and finalization.
In the first phase, the server sends the announcement of a predetermined multicast address in
advance to indicate the start of the shipment. Send information file sizes, etc. The clients send
information about they are prepared to receive and the server responds with confirmation
that the message has arrived.
In the second phase, the sender numbers and sends packets by UDP multicast, at the end of
the transmission sends a completion message and the receivers respond with a list of packages
that have been misused. The sender sends the packets that are not ACK and repeats the cycle.
In the finalization, when customers get the file right, send a completion message to the server,
and when receives a message from server announcing that he has reached the completion
message; it stay waiting in the multicast address for new shipments.
1.3.5 UNICAST / MULTICAST TRANSMITION
A unicast transmission is performed when the data stream is sent from one terminal to
another, so that when a user connects to a service, it is created another data stream to his
terminal. There is a data stream for each terminal that connects to the server. In illustration
4-9 it can be seen how it is sent a different stream for each computer requesting the live
video that the camera is capturing.
ILLUSTRATION 4-9 UNICAST TRANSMISSION
Multicast transmission involves sending a single stream to all terminals that want the camera
image. Multicast transmission can be of two types: Sparse and dense. Sparse mode generates
traffic across the entire network even if the terminals do not want to see the image, as shown
in Part A of illustration 4-10. The dense mode, traffic is generated only if the terminal requests
the service (illustration 4-10 B).
ILLUSTRATION 4-10 MULTICAST TRANSMISSION
The network cards or NICs (network interface controller) in the terminals or communications
equipment of a LAN segment, only receive packets with its MAC address or MAC address
broadcast. The LAN IEEE 802.3 standard provides the ability to support multicast.
The first bit of link-level frame is the one that indicates whether the frame is multicast or
unicast, If is 1 is multicast. In this way the networks equipments at level 2, knows that is a
multicast message and forward the packet they are receiving. This help to have lower network
latencies, as can be seen in illustration 4-11.
ILLUSTRATION 4-11 MULTICAST ETHERNET FRAME
To support multicast transmission at the network level, the Internet Assigned Numbers
Authority (IANA) has set a series of addresses for these transmissions (220.127.116.11 to
18.104.22.168), so when a router receives a message with a these addresses, known is to be
multicast and forwards by all segments of your network (Sparse mode) or segments have a
terminal that has been associated with a transmission (dense mode).
Para soportar la transmisión multicast a nivel de red, el Internet Assigned Numbers Authority
(IANA), ha establecido una serie de direcciones para estas trasmisiones (22.214.171.124 a
126.96.36.199), de esta forma cuando un encaminador recibe un mensaje con una de estas
direcciones, sabe que es multicast y lo reenvía por todos los segmentos de su red (modo
esparcido) o en los segmentos que haya algún terminal que se haya asociado a una
transmisión (modo denso).
To determine when a router has to transmit the multicast packets in the packet level three,
there is a field that is Time To Live (TTL), this field indicates the number of routers through
which can pass the package before it was removed (not broadcast).
Each time a packet arrives to the router. It looks at the TTL field and if it is 1 do not broadcast
and if it is greater decreases in 1 and transmits. This avoids sending a packet to be going
around the world, from router in router. The values are 1 for local, 15 for metropolitan areas,
63 and 127 wan environments for broadcast worldwide.
The network management is performed by heterogeneous firms; some companies will not
allow multicast traffic circulation or transit through them of multicast traffic. To overcome this
and to connect multicast networks linked by networks or links that do not support multicast,
there is performed IP tunnels on which the datagrams are encapsulated in unicast datagrams
(point to point) within the network, and at the output of the network, they are switch back to
1.3.6 INTERNET GROUP MANAGEMENT PROTOCOL (IGMP)
The Internet Group Management Protocol (IGMP) IETF Standard (RFC 1112) is the
recommendation used by routers to learn of the existence of members of groups (MC)
connected to their subnetworks. The operation is carried over UDP, and has two kinds of
Host Membership Query (HMQ)
Host Membership Report (HMR)
Each router periodically sends a message to Layer 2 multicast IGMP HMQ (188.8.131.52)
requesting information to hosts on the LAN, the message is sent with a TTL = 1 so that does
not leave the LAN environment.
The computer that has a multicast service sends a response message for each group. If an
application wants to subscribe to a multicast group, the driver creates a new hardware MAC
address a message to the router.
The routers send IGMP information to other routers to propagate information.
Once established services, for each service it is created a pair (source, target group), multicast
traffic is transmitted over the network using spanning tree by connecting all the computers in
In dense mode, when there are many users within the network, is performed by spanning tree
protocol variations, such as: Distance Vector Multicast Routing Protocol (DVMRP), Multicast
Open Shortest Path First (MOSPF) or Protocol‐Independent Multicast‐Dense Mode (PIM‐ DM)..
To Sparse mode, when there are few users in the network, more selective techniques are used
to avoid wasting bandwidth such as Core Based Trees protocol (CBT) or Protocol-Independent-
Sparse Mode (PIM-SM)
When a packet arrives at a router, it has to remove the wrapper from level 2 to treating the
head of level 3, process, modify, re-wrap it with the plot of level 2 and broadcast by the
network. These tasks involve a cost in time and processing capacity.
The mechanism of Multiprotocol Label Switching (MPLS) was developed by IETF to expedite
the transport of IP datagrams in the operator’s backbone networks and reduce the processing
of a datagram in the nodes, changing routing for switching.
It works between levels two and three introducing a label similar to a virtual circuit, so the
datagrams are switched instead of routed. Takes advantage of the fact that connective
networks once are connected, are faster than non-connective.
The entry node, insert a tag that will be used for the packet to circulate for a preset route. The
last node removes the tag.
The process into a router is:
Extract the destination address of the datagram, at level 2.
Search into the routing table
It is not a mechanism that has been created for streaming media. It was created to reduce the
delay in the networks. This decrease is an advantage in the transmission of picture and sound,
especially when this transfer involves a conversational transmission.
The protocol suite RTP / RTCP (Real‐time Transport Protocol / Real‐time Transport Control
Protocol) defines mechanisms, equipment and data packets for transmission of audio or video
in a unidirectional or bidirectional, unicast or multicast mode.
The protocol does not guarantee:
Recovery from errors or packet loss during transmission (integrity).
The quality of the material.
That packet arrives in order.
The availability of bandwidth required for applications (QoS).
It consists basically of two distinct parts:
The real-time protocol at the transport layer, to transport traffic labeled real-
The RTP Control Protocol (RTCP), to monitor the quality of service and transmission
of information to users who are online.
The RTP transport protocol is designed for communications that require data transfer in real
time over the Internet, including control mechanisms for synchronizing different streams of
material. The basic elements to deal with considered real-time traffic are:
Time code associated with the data. Indicates the cadence with which it is
necessary to reproduce data.
Packet sequence number. To indicate the order of packets, does not guarantee that they
arrive in order, not even all reaching their destination, it serves to indicate the playback
sequence, the receiver will have to order them for reproduction and in the absence of
packets, to perform statistics or accommodate the flow to the line
ILLUSTRATION 4-12 PROTOCOLS STACK FOR LEVELS 3, 4 AND 5
In illustration 4-12 it can be see where it rests each of the two parts of the recommendation.
RTP is always through UDP transmissions, while RTCP can be implemented by either of the two
protocols, if it does not want to lose RTCP packets are sent over TCP and if the packet loss is
allowed, it can be transmitted using UDP.
In the beginning was designed for audio or video transactions using this basis:
Users are grouped in one address
The data of audio / video is chopped into packets (20mS).
Each user can have different encoding parameters.
Header with time code and packet number.
Each service is on a different port (+1 for RTCP).
Audio and video separately on different ports.
Line quality -> packages lost, order arrival, jitter.
Ports associated with services (audio, video).
In illustration 4-13 you can see when the protocol comes into play, once the communication is
established over HTTP connections, it can start the images and sound transfer, the server
sends packets using RTP over UDP and RTCP sends packets from the receiver to the sender to
send data such as: delay, packet loss, jitter, etc.
ILLUSTRATION 4-13 PACKET EXCHANGE RTP/RTCP.
The recommendation specifies intermediate equipment that may be between the source and
Mixers - Due to the heterogeneity of networks, it is possible that one of the
users of the audio or video conferencing cannot absorb the traffic generated by
various video and audio connections and therefore do not have to be necessary
for all work to lower quality. To fix this, the packets that reach people of lower
quality in many streams are played back on a stream and re-encode at a lower
speed in order to be received by the user. In Illustration 4-13 you can see two
signals in two different streams are mixed in a single image and encoded into a
Translator. Other cases that are specified in the standard is the possibility that a user has
the conference system within a local network and adapted to this speed, the problem
that arises is that the Internet output line cannot handle the data flow with the quality
generated, so that at the exit to the Internet will be necessary to lower the transmission
rate in a similar way to the mixers
Layer codification. Multimedia applications need to adjust his speed between the
transmitter and receiver for a proper reception of the supplied material, when only are
two devices, it is relatively simple, the problem resides when is multicast in which it
cannot force most users to see the quality of the broadcast to the lower speed of the
users. For this type of transmission is defined certain layers with different coding rates
so that each terminal will "engage" with the service best suited to their characteristics.
ILLUSTRATION 4-14 MIXTURE OF TWO IMAGES IN A SINGLE.
1.3.9 RTSP (REAL-TIME STREAMING PROTOCOL).
RTSP is a recommendation that establishes the mechanisms necessary for the remote control
of video servers, for this control, it specified a series of commands, they are:
OPTIONS: Request of available commands.
SETUP session establishment
ANNOUNCE change the description of the media object
DESCRIBE, description of the object
REDIRECT to redirect to another server
SET PARAMETER change the encoding parameters or device
As an extension of Illustration 4-14, illustration 4-15 shows in which it is integrate RTSP
commands, after logging HTTP, RTSP commands are sent to indicate the actions to be
undertaken by the video server, the audiovisual material is sent via RTP.
ILLUSTRATION 4-15 RTP, RTCP Y RTSP COMMANDS INTERACTION
In illustration 4-16 it can see where it is located RSVP, as well as RTCP can work over TCP and
UDP and audiovisual material is on RTP and UDP.
ILLUSTRATION 4-16 LEVELS 3, 4 Y 5
1.4 QUALITY OF SERVICE
The protocols that have been seen before, they try to take the most of the line they have, but
do not guarantee any quality on the line.
The Quality of Service (Qos), is the ability of networks to ensure, with any degree of reliability,
that the transmission of information will meet certain parameters. QoS enables provide better
service to users and better use of the network. In the case of IP networks would be the
transfer of "packets" to meet certain parameters.
QoS parameters that are most interested in video streaming over IP networks are:
Delay. In the case of needing to work in conversational mode.
Jitter. So that it can play the video without stopping.
Error Rate (losses). To have the greatest possible data integrity.
Capacity (bandwidth). To ensure a certain quality in the picture.
Class of Service (CoS). Not all traffic is to have the same restrictions in terms of QoS.
Groupings of various types of traffic which is assigned a certain priority, a common
treatment in the network nodes that have QoS capabilities.
A communications provider offers a "Quality of Service" when it guarantees the value of one
or more parameters. When the provider does not guarantee any QoS parameter, then there is
not QoS and it is called: best effort service (best effort). Internet and IP networks in general
have been designed to offer a single type of service: "best effort".
The service "best effort" does not guarantee QoS, especially when the network is congested.
Generically we can say that in these networks the QoS parameters vary at random,
unpredictable, and may exceed allowable limits for different types of applications.
Mechanisms are needed to ensure QoS in IP networks.
Network models that provide quality of service are: Integrated Services (IntServ),
Differentiated Services (DiffServ) and MPLS. IntServ and DiffServ are models for quality of
service at the IP level (3). These services have in common the existence of border and core
The border routers are on the edges of the network and process incoming traffic, characterize
and penalize the user’s traffic and may reject requests for QoS of users. The processing core
routers traffic from other routers (core or border) differentiating the traffic you need to tackle
1.4.1 INTSERV Y RSVP
The IntServ (Integrated Services Architecture, ISA) began his definition in 1994 with the idea of
expanding the existing IP architecture to support real-time traffic with QoS, while maintaining
best-effort service. IntServ is a global architecture for QoS over best effort IP network that
supports two types of services (in addition to best effort):
Controlled Load (CL)
Guaranteed Service (GS)
IntServ is designed for individual applications, and quality of service is achieved by:
Presentation to the network of application requirements.
Application of a certain quality of service.
Monitor the status of each flow in each network node is required from non-permanent
(Soft States) on each node IntServ
IntServ is a network model for flow-based QoS and resource reservation in advance. The
IntServ flow is the "set of packages ranging from one source to one or more destinations, and
receive a common treatment for QoS" can also be defined as "the stream of packets only
result of a user's activity, which requires the same QoS". Is always treated as one-way traffic
An IP packet identifies to belong to an IntServ flow (flow ISA) by:
Source and destination IP address.
Source and destination port number
Transport protocol used (TCP o UDP)
Nodes reserve resources between a source and a destination, so that an "IntServ" stream can
be considered as a kind of "virtual circuit”, it will require signage for: establishment,
maintenance and release.
How to implement a model based on IP flow, when IP is a no connective protocol? It uses a
signalling protocol over IP as a mechanism to reserve resources at each node along the route
between origin and destination: RSVP for signalling the requirements of bandwidth and latency
RSVP (ReSerVation Protocol resorces). It is a resource reservation protocol by default for the
IntServ model. In this model, resources are space in the buffers of routers and bandwidth on
the links. RSVP offers a set of procedures that enable applications and nodes (routers) to
communicate their QoS requirements enabling the processing of data streams with QoS
from previous reservations. It is a independent protocol from IntServ model and it can be
used in other models or architectures.
Resource reservation, by RSVP terminology can be performed over unicast communications
when there is point to point and multipoint to point. For Multicast connections are point to
multipoint and multipoint to multipoint.
Diffserv (differentiated services) allows differentiation and scalability of services in IP networks
and the Internet. DiffServ is based on separate in the network nodes the forming and
forwarding functions. Packet forwarding is based on "per-hop behavior (PHB) and the
differential treatment of each packet in a network node based on queue disciplines, which is
NOT part of the Diffserv specification. Not based on: sign or reservation resource. DiffServ
routers do not see the "flows."
Diffserv enables interoperability with Non Diffserv nodes showing flexibility for services, it
does not define specific services or services classes and work in a contrary manner to IntServ
as it provides functional "pieces" of a network architecture over to build services, allowing
flexibility before the emergence of new classes or obsolescence of other already defined.
The recommendation contains many aspects and is based on the management of queues at
Classifies packets into categories (by type of service requested).
Each category has a SLA (Service Level Agreement). It can be hire a given flow rate in
a desired category.
The routers process each packet according to its category.
The Policy/Admission Control is performed only in the entry routers to the provider's
network and between different providers.
Common kinds of service are:
Kinds of services:
“Premium” ó “Expedited Forwarding” RFC2598. Similar to a leased line.
Guarantees: jitter, delay and throughput.
‘”Assured Forwarding” RFC2597. Preferential treatment with no SLA, no guarantees.
Four classes defined three priorities each.
Best Effort (with priority). No guarantee but preferential treatment.
Best Effort (without priority). No warranty, waste is used.
For this kind of Quality of Service, it can be used a queue management in the nodes as shown
in illustration 4-17. When a packet arrives with the upper category, is placed in the "expedited"
queue, messages in this queue are broadcast as they come. Each node has implemented a
number of queues for each one of the kinds of qualities of service offered. In the example of
the illustration are implemented six qualities of service in the node.
ILLUSTRATION 4-17 QUEUE PLANNER FOR DIFFSERV
1.5 FEC (FORWARD ERROR CORRECTION)
Even in safer lines, there are often errors in transmission; these errors require retransmissions
that depending on the type of network are made between links or among terminals. In video
transmission, that use to be in real time, these retransmissions are not useful because when
the information arrives, has already passed the time that ought to be.
To avoid retransmissions caused by errors in transmission lines, created the FEC mechanism,
consists in to add redundancy to the information to be transmitted so that the receiver can
correct transmission errors. The more redundancy is added in the transmission, the greater the
number of errors that can occur in a transmitted packet, but the lower the percentage of
capacity utilization of the link.
The kinds of FEC used in the links are divided into: block codes and convolutional codes. Block
codes consist of applying one or more expressions that generate the redundancy of a full
block, the reception is applied for reverse to reconstruct the original block damaged. In the
case of convolution, the encoding is done character by character.
1.5.1 SMPTE 2022‐2007
As can be seen, there are several methods to provide quality of service in networks so that
data from audio / video, have a higher priority of other data. These resource stocks do not
preclude that there may be congestion at times given or failures of transmission lines that
cannot be corrected by FEC codes from the links. These circumstances can make to lose a
package or if there is an incorrect part, then make the whole package unusable, regardless the
need to perform a transmission over a network without quality of service so the problem
Thus there are two forms of packet loss: the erroneous packets that the link layer cannot
reconstruct by the number of errors and especially the eliminated packages by network
congestion. Depending on the type of data within the packet are wrong, the consequences of
their loss can be disastrous, including the inability to play.
To solve this problem, the Video Services Forum (VSF) published the SMPTE 2022-2007
recommendation which extends the same philosophy that FEC made on the bytes to try to
reconstruct the erroneous bytes in packets to try to reconstruct lost packets.
The idea is to fill, by rows, an array of LxD RTP packets while the transmission is made and
when the matrix is filled, it creates a redundancy packet for each column as shown illustration
ILLUSTRATION 4-18 FEC PACKET GENERATION
The system works as follows. In illustration 4-19 it can see a 5x5 system as 5 redundant packets
that are generated per 25 packets of useful information. To recipient the RTP packets will
arrive in order. On the transmitter from generation of packet 21 it can be generated the C1
redundancy packet. Suppose the packet number 12, the network has ruled out any reason, the
receiver will have to wait for the arrive of C1 redundancy package to generate the packet 12
from the packages 2, 7, 17, 22 and C1.
ILLUSTRATION 4-19 ERROR IN ONE PACKET
It should be noted that the redundancy is performed by columns, because the locks or
saturation of the network are usually close in time, so if it has ruled packet 12 for saturation,
packet 13 or 14 will be more likely to be discarded that 17.
Through this system we can have a blast of discards or error in the number of columns (just
can has one error per column).
1.5.2 FEC 2D
The problem comes when they fail two packets of the same column as shown in illustration 4-
20. The receiver cannot reconstruct the packets 12 and 22 from 2, 7, 17 and C1.
ILLUSTRATION 4-20 TWO ERRORS IN A COLUMN
To remedy this situation, the Video Services Forum published 2D SMPTE 2022-2007 - FEC,
consisting of the creation of redundancy in the rows as shown in illustration 4-21. With this
system we have several erroneous packets in the same column.
ILLUSTRATION 4-21 2D FEC