Introduction
to Asynchronous Transfer Mode
This chapter briefly describes and
explains the key features, functions, and benefits of asynchronous transfer
mode networking, and discusses how Microsoft® NetShow® Theater
Server operates on this network type. This information is presented in
the following topics:
Understanding
Asynchronous Transfer Mode
Asynchronous transfer mode (ATM)
is a digital packet-switching technology originally developed to integrate
voice, video and data traffic into one network. An ATM network is a connection-oriented
environment that easily can be optimized for video-on-demand delivery systems.
It offers extremely high throughput and guaranteed transmission of sequential
packets.
ATM uses very short, fixed-length
packets called cells. Each cell has a fixed length of 53 bytes. By using
fixed-length cells, the information can be transported in a predictable
manner. The ATM cell is broken into two main sections, the header and the
payload. The header (5 bytes) is the addressing mechanism, and is significant
for networking purposes because it defines how the cell is to be delivered.
The payload (48 bytes) is the portion that carries the actual voice, data,
or video information.
For information on configuring
your ATM network or configuring NetShow Theater Server for an ATM network,
see Microsoft® NetShow Theater Server Installation Guide
Understanding
the OSI Reference Model
ATM networks are described in a
layered architecture defined by the Open Systems Interconnection (OSI)
reference model, the International Standards Organization (ISO) structure
for the "ideal" network architecture. This model outlines seven areas,
or layers, for the network. These layers (from highest to lowest) are:
-
Applications. Where the user
applications software lies. Such issues as file access and transfer, virtual
terminal emulation, interprocess communication, and the like are handled
here.
-
Presentation. Differences
in data representation are dealt with at this level. For example, UNIX-style
line endings (CR only) might be converted to MS-DOS style (CRLF), or EBCIDIC
to ASCII character sets.
-
Session. Communications between
applications across a network is controlled at the session layer. Testing
for out-of-sequence packets and handling two-way communication are done
here.
-
Transport. This layer makes
sure the lower three layers are operating correctly, and provides a transparent,
logical data stream between the user and the network service being used.
This is the layer that provides local user services.
-
Network. This layer makes
certain that a packet sent from one device to another actually gets there
in reasonable time. Routing and flow control are performed here. This is
the lowest layer of the OSI model that can remain ignorant of the physical
network.
-
Data link. The purpose of
this layer is to manage the transmission of data packets between the network
layer of the source and destination computers. This layer also provides
error detection and correction, and retransmission. Generally, this layer
is broken into two sublayers: The Logical Link Control (LLC) on the upper
half, which does the error checking, and the Medium Access Control (MAC)
on the lower half, which deals with getting the data on and off the wire.
-
Physical. This is the hardware
layer. Here is where the cable, connector and signaling specifications
are defined.
The OSI reference model is a generalized
description of networking functions. It provides a common set of terms
that can be related to a specific networking system. If you understand
how one network operates and understand how that network is related to
the OSI model, then you can quickly understand a new network by seeing
how that network also relates to the OSI model.
Understanding
ATM Network Architecture
Asynchronous transfer mode (ATM)
layering is made up of three basic layers: the physical layer, the ATM
layer, and the ATM adaptation layer. The following table relates the general
Open Systems Interconnection (OSI) reference to the layers and sublayers
of ATM networks.
| OSI |
ATM |
| Transport/Network |
AAL |
CS SAR |
Convergence Segmentation and reassembly |
| Network/Datalink |
ATM |
|
Generic flow control Cell VPI / VCI translation Cell multiplex
and demultiplex |
| Physical |
PHY |
TC |
Cell rate decoupling Header Error Control (HEC) Cell delineation
Transmission frame adaptation Transmission frame generation |
| |
|
PM |
Bit timing Physical medium |
In this table: AAL is the ATM
adaptation layer, CS is the convergence sublayer, SAR is the segmentation
and reassembly sublayer, ATM is the ATM layer, PHY is the physical layer,
TC is the transmission convergence sublayer, and PM is the physical
medium sublayer.
About Virtual
Paths and Virtual Circuits
On a connection-oriented network,
the routing between the endpoints of the connection is negotiated when
the connection is established. This routing persists as a service reservation
for the duration of the connection. This ensures cell sequence at point
of reception, and also allows the negotiation of the quality of service
transmission parameters that optimize the connection for video-on-demand
service.
Each cell contains a connection
identifier in its header that uniquely identifies the connection throughout
the network. Two such hierarchical identifiers are located in the header
of each cell: a virtual path (VP) and a virtual circuit (VC). A virtual
path is the generic term for a collection of virtual circuit links. A virtual
circuit is a unidirectional transport of ATM cells. The VC identifier (VCI)
identifies a particular VC link for a given VP connection.
Organization of the cell frame
header
Suppose a virtual circuit consisting
of a four-hop path is selected between users A and B by a routing algorithm.
After the network finds a path between the two nodes, it assigns the VCI
values to be used at each node along the path and sets up routing tables
at nodes N1 to N5. When the transmission starts, all cells of the connection
follow the same path in the network. Once the communication is completed,
one of the two users releases the connection, and the VC also is terminated.
The end-to-end connection defined by the concatenation of VC links is called
a virtual circuit connection (VCC).
About the Physical
Layer
The physical layer of the ATM network
takes care of two separate classes of functions:
-
Those that relate to the physical
medium (PM) used
-
Those that change bits to ATM cells:
transmission convergence (TC)
The first task for any transmission
system is to synchronize timing at the bit level. This is 155.52 megabits
per second (Mbps) for ATM systems operating over optical fiber. Once the
bits are available to the next sublayer, it is possible first to convert
the bits to the frames of the transmission system used, and then to convert
the frames to the actual cells.
The transmission convergence
sublayer is standardized to generate and extract frames at the rates specified
from the transmission system. The ATM cells are found by looking for the
header error control (HEC) on the cell header and checking the error correction
code. Once the cells are checked, they must be decoupled from the transmission
rate of the physical medium by inserting active cells into and deleting
idle cells from the stream. These cells then are handed off to the ATM
layer.
About ATM Layer
The ATM layer deals with cells and
their transport; it defines cell layout and the meaning of header fields.
The ATM layer includes two interfaces: the user-network interface (UNI)
and the network-network interface (NNI). The UNI defines the boundary between
a host and an ATM network (or the customer and the carrier). The NNI refers
to the line between two ATM switches. The header field varies slightly
between the UNI and NNI.
The ATM layer routes cells across
the network and manages the transmission of cells together from many virtual
paths onto one physical carrier. It also is responsible for delivery of
a quality of service to the higher layers. The cell loss priority (CLP)
attribute designates whether a cell is high or low priority. The payload
type (PT) indicator defines the file type of the cell. The header error
control (HEC) prevents errors from occurring in the header itself.
About ATM Adaptation
Layer
The ATM Adaptation Layer (AAL) is
used to adapt the ATM layer to the services that are using the network.
It has two sublayers: the convergence sublayer (CS) and the segmentation
and reassembly sublayer (SAR). The CS converts user service information
into a protocol data unit (PDU). The SAR places these PDUs into cells,
resulting in a 48-byte SAR-protocol data unit with a 5-byte header that
together make up the 53-byte cell.
There are four different classes
of service delivered by the AAL based on timing between source and destination,
the bit rate of the source, and the connection mode. The following diagram
describes how these service classes are defined.
Functional definition of ATM
service classes
All Microsoft® NetShow Theater
Server services function over AAL5. The AAL5 format standardizes how the
cells are generated and which headers they include.
Understanding
ATM Services
Asynchronous transfer mode (ATM)
provides quality of service (QoS) to connections. QoS is a set of parameters
specified by the user service, such as delivery time and bandwidth. While
setting up the connection, the service and the network negotiate the QoS
parameters to govern that connection. Microsoft® NetShow Theater Server
video streaming traffic is in AAL5 format, and a specified Constant Bit
rate (CBR) service. CBR service specifies the mean bit rate and cell loss
acceptable, as well as the allowable variation in the cell delay.
Each time a new connection is
opened, ATM Connection Admission Control (CAC) reviews the current demands
on the network, evaluates the required service against the remaining network
resources available, and accepts the new connection only if sufficient
resources are available. In addition, ATM switches are configured to accept
new traffic only up to a maximum percentage (typically 90 percent) of their
theoretical capacity.
About the Network
Perspective
ATM networks scale effectively in
both local area network (LAN) and wide area network (WAN) environments.
Network control functions can be automated or distributed efficiently.
Network control and management functions try to satisfy three competing
objectives:
Balancing competing functions
in network management
NetShow Theater Server assumes
a dedicated networking environment. For example, NetShow Theater Server
Setup reports the calculated maximum number of streams supported. This
assumes that there is no competing network traffic that would reduce the
available network resources.
Understanding
ATM Connection Setup
Asynchronous transfer mode (ATM)
connections are established as either permanent virtual circuits (PVCs)
or switched virtual circuits (SVCs). As their name implies, PVCs are always
present, whereas SVCs must be established each time a connection is set
up.
To set up a connection, a signaling
circuit is used first. This is a predefined circuit (with VPI equal to
0 and VCI equal to 5) that is used to transfer signaling messages, which
are used for making and releasing calls or connections. If a connection
request is successful, a new set of VPI and VCI values are allocated on
which the parties that set up the call can send and receive data.
Six message types are used to
establish virtual circuits. Each message occupies one or more cells, and
contains the message type, length, and parameters.
| Message |
Significance if sent by host |
Significance if sent by network |
| SETUP |
Request to establish a call |
Indicates an incoming call |
| CALL PROCEEDING |
Acknowledges an incoming call |
Indicates a call request will be attempted |
| CONNECT |
Indicates acceptance of a call |
Indicates a call was accepted |
| CONNECT ACK |
Acknowledges acceptance of a call |
Acknowledges making a call |
| RELEASE |
Requests that a call be terminated |
Terminates a call |
| RELEASE ACK |
Acknowledges releasing a call |
Acknowledges releasing a call |
The sequence for establishing and
releasing a call is:
-
The host sends a SETUP message on
the signaling circuit.
-
The network responds by sending
a CALL PROCEEDING message to acknowledge receiving the request.
-
Along the route to the destination,
each switch receiving the SETUP message acknowledges it by sending the
CALL PROCEEDING message.
-
When the SETUP message reaches its
final destination, the receiving host responds by sending the CONNECT message
to accept the call.
-
Next, the network sends a CONNECT
ACK message to acknowledge receiving the CONNECT message.
-
Along the route back to the sender,
each switch that receives the CONNECT message acknowledges it by sending
CONNECT ACK.
-
To terminate the call, a host (either
the caller or the receiver) sends a RELEASE message. This causes the message
to propagate to the other end of the connection, and then releases the
circuit. Again, the message is acknowledged at each switch along the way.
About Sending
Data to Multiple Receivers
In ATM networks, users can set up
point-to-multipoint (P/MP) calls, with one sender and multiple receivers.
A P/MP VC allows an endpoint called the root node to exchange data
with a set of remote endpoints called leaves. To set up a point-to-multipoint
call, a connection to one of the destinations is set up in the usual way.
Once the connection is established, users can send the ADD PARTY message
to attach a second destination to the VC returned by the previous call.
To add receivers, users then can send additional ADD PARTY messages.
This process is similar to a
user dialing multiple parties to set up a telephone conference call. One
difference is that an ATM P/MP call doesn't allow data to be sent by parties
toward the root (or the originator of the call). This is because the ATM
Forum Standard UNI 3.1 only allows data on P/MP VCs to flow from the root
toward the leaves.
Understanding
ATM Switching
An asynchronous transfer mode (ATM)
switch transports cells from the incoming links to the outgoing links,
using information from the cell header and information stored at the switching
node by the connection setup procedure. The connection setup procedure
performs the following tasks:
-
It defines a unique connection identifier
(a virtual path identifier [VPI] and virtual channel identifier [VCI])
for each connection at the incoming link and link identifier, and a unique
connection identifier at the outgoing link.
-
It sets up routing tables at each
switching node to provide an association between the incoming and outgoing
links for each connection.
As mentioned previously, the VPI
and VCI are the connection identifiers used in ATM cells. To uniquely identify
each connection, VPIs are defined at each link and VCIs are defined at
each virtual path (VP). To establish an end-to-end connection, a path from
source to destination must be determined first. Once the path is established,
so is the sequence of links to be used for the connection and their identifiers.
VPIs are used to reduce the amount
of processing required at an ATM switch by routing on the VPI field only.
For example, VPI routing is useful when many VCs share a common physical
route (similar to all phone connections between Seattle, WA, and Chicago,
IL).
Understanding
ATM Service Types
Asynchronous transfer mode (ATM)
standards define the types of services most commonly used, in order to
control the various categories of network traffic. The service categories
are listed below.
About Constant
Bit Rate
Constant Bit Rate (CBR) services
generate traffic at a constant rate. With this class of service, the end
station must specify bandwidth at the time a network connection is established.
The network then commits that bandwidth along the connection route; this
ensures that no network traffic is lost due to traffic congestion because
connections are admitted only if the requested bandwidth can be guaranteed.
Microsoft® NetShow~Y Theater Server requires CBR.
About Constant
Bit Rate
Constant Bit Rate (CBR) services
generate traffic at a constant rate. With this class of service, the end
station must specify bandwidth at the time a network connection is established.
The network then commits that bandwidth along the connection route; this
ensures that no network traffic is lost due to traffic congestion because
connections are admitted only if the requested bandwidth can be guaranteed.
Microsoft® NetShow~Y Theater Server requires CBR.
About Variable
Bit Rate
Variable Bit Rate (VBR) is divided
into two subclasses, real-time VBR (RT-VBR), and Non-Real-Time VBR (NRT-VBR).
RT-VBR is for applications that have variable bit rates and stringent real-time
requirements, such as interactive compressed video (video-conferencing,
for example). Real-time services can deteriorate in quality, or become
unintelligible if the associated information is delayed; they are sensitive
to the time it takes for the ATM cells to be transferred.
NRT-VBR is for traffic where
timely delivery is not as important~Wthat is, the quality of non-real time
services is not affected by delays in information transfer. An example
of non-real-time service is data transmission.
About Available
Bit Rate
Available Bit Rate (ABR) service
is for bursty traffic for which an approximate bandwidth is known. Burstiness
is the ratio of peak-to-average traffic generation rate. With ABR, you
can specify, for example, a fixed capacity of 5 megabits per second (Mbps)
between two points, with peaks of up to 10 Mbps. The network guarantees
5 Mbps will be provided at all times, and tries to provide the peak capacity
when needed, but does not guarantee this.
With ABR service, the network
provides feedback to the sender, asking it to slow down when traffic congestion
occurs.
About Unspecified
Bit Rate
Unspecified Bit Rate (UBR) service
allows a connection to be established without specifying the bandwidth
expected from the connection. The network does not guarantee UBR service;
it establishes the route, but does not commit bandwidth. UBR service can
be used for applications that have no delivery constraints, and do their
own error and flow control, such as e-mail and file transfer, which have
no real-time characteristics.
Understanding
ATM Traffic Control
Traffic control refers to a set
of actions performed by the network to avoid congestion, and to achieve
predefined network performance objectives, for example, in terms of cell
transfer delays or cell loss probability.
The International Telecommunications
Union (ITU, formerly the Consultative Committee on International Telegraphy
and Telephony) has defined a standard rule, called the Generic Cell Rate
Algorithm (GCRA) to define the traffic parameters. This rule is used to
differentiate clearly between conforming and nonconforming cells~Wthat
is, it provides a formal definition of traffic conformance to the negotiated
traffic parameters.
ITU recommendation I.371 defines
two equivalent versions of the Generic Cell Rate Algorithm: the virtual
scheduling (VS) algorithm, and continuous-state leaky bucket (LB) algorithm.
For any sequence of cell arrival times, both algorithms determine the same
cells to be conforming or nonconforming.
GCRA uses the following parameters:
-
AT is the cell arrival time.
-
I is the increment, or the distance
between two consecutive cells.
-
L is the limit, representing a certain
tolerance value.
-
TAT is the theoretically predicted
cell arrival time.
When a cell arrives, the VS algorithm
calculates the TAT of the cell, assuming the cells are equally spaced when
the source is active. If the AT of a cell is equal to TAT - L, then
the cell is conforming; otherwise, the cell arrived too early and is considered
nonconforming.
The continuous-state leaky bucket
algorithm can be viewed as a finite capacity bucket whose contents leak
out at a continuous rate of 1 per time unit and whose contents are increased
by I for each conforming cell. If the contents of the bucket are less than
L when a cell arrives, then the cell is conforming; otherwise, it is nonconforming.
The capacity of the bucket equals L + I.
For more information about the
CGRA algorithm, see ITU recommendation I.371.
About Traffic
Control Objectives
Asynchronous transfer mode (ATM)
traffic control has the following objectives:
-
To support a set of quality of service
(QoS) classes that are sufficient for existing and foreseeable services.
-
To maximize network resource utilization.
-
To use network resources efficiently
under any traffic conditions.
Two basic traffic control functions
that ATM networks use to manage traffic are connection admission control
(CAC) and usage parameter control (UPC).
About Connection
Admission Control
Connection admission control (CAC)
represents a set of actions carried out by the network during the call
setup phase to accept or reject an ATM connection. If sufficient resources
exist to accept the call request, and if the call assignment does not affect
the performance quality of existing network services, then the call is
granted. At call setup time, the user negotiates with the network to select
the desired traffic characteristics.
About Usage Parameter
Control and Network Parameter Control
Using usage parameter control (UPC)
and network parameter control (NPC), the network monitors user traffic
volume and cell path validity. It monitors users~R traffic parameters to
ensure that they do not exceed the values negotiated at call setup time.
It also monitors all connections crossing the user-network interface (UNI)
or network-network interface (NNI).
The UPC algorithm must be capable
of monitoring illegal traffic conditions, determining whether or not the
confirmed parameters exceed the limits of the negotiated range, and quickly
dealing with parameter usage violations. To deal with such violations,
the network can apply several measures, for example, discarding the cells
in question, or removing the connection that contains those cells.
About the Priority
Control Function
The Priority Control (PC) function
is used to support the actions of CAC and UPC/NPC. Users can employ the
cell loss priority (CLP) bit to create traffic flows of different priorities,
allowing the network to selectively discard cells with low priority if
necessary to protect those with high priority.
About Network
Resource Management
Network resource management (NRM)
represents provisions made to allocate network resources to separate network
traffic flows according to service characteristics.
© 1996-1998 Microsoft Corporation. All rights reserved.