ATM applications and quality of service

Introduction
There are many network parameters that need to be configured to deliver an ATM service. The flexibility of the technology combined with the large number of variables makes ATM services difficult for end users to interpret. This also leads to difficulty in proving the business case for adopting ATM services.

This webpage is my attempt to provide high level definitions of suggested ATM services combined with typical applications of these services to better understand and to facilitate the adoption of ATM technology as a business solution.

Quality of service
Quality of service (QoS) is a measure of performance of a network system that reflects its transmission quality and service availability.

Quality of service parameters in ATM technology revolve around a combination of the following time delay and cell loss attributes.

These parameters are too complex for most end users to define to an ATM service provider for a particular application. A better approach is for the ATM service provider to offer the end users applications examples with some predefined parameters for each ATM QoS class.
The services discussed in will include a description of the applications and the likely uses of the service and will also include discussions on the QoS class and its relation to traffic contracts; circuit priority; PVCs and SVCs; point-to-point and multipoint circuits.
ATM is a flexible networking technology, it supports several traffic types as shown in Table 1.
Quality of Service (QoS) Class Timing Delay & Cell Loss Parameters Attribute Typical Applications (Traffic type)
CBR Stringent Voice, circuit emulation
VBR-RT Strict Real-time video and audio
VBR-NRT Average Broadcast video
Mission critical data, frame relay interworking
ABR, UBR Low Bursty data traffic, LANs
Table 1.
Traffic Supported by various QoS classes


ATM services description
1- ATM data PVCs
Service description
One of the simplest ATM service offerings is an ATM data PVC service. This offering is very similar to Frame Relay. The primary difference is that access speeds are higher (E3 or OC3) and ATM is the method of access.
An ATM data PVC service offering is depicted in Figure 1.
ATM data PVCs
Figure 1
ATM data PVCs

This service is targeted at connecting routers at speeds of E3 and above in a PVC network. There are multiple ways to offer ATM data PVC services using different QoS classes. In this particular case it is suggested that two types of basic services are offered to the end-user:

The best effort delivery uses the UBR service class. This provides large buffers and uses bandwidth that is not reserved by other QoS classes. Best effort service will allow bursting of cells up to the access rate and will deliver cells as long as none of the queues along the path overflow.
On the other hand a VBR-NRT service provides users with PVCs and associated traffic contracts. This service is similar in nature to Frame Relay services at higher speeds.
The primary difference is that the traffic contract is more complex than Frame Relay in that it clearly defines peak cell rates (PCR) and maximum burst sizes (MBS) as well as average throughput or sustained cell rate (SCR).
Also, ATM CPE (customer premises equipment) devices that are connected to a VBR-NRT service must implement traffic shaping to maintain the connection to the network.

Traffic contract
The best effort service does not have a traffic contract. In the ATM traffic descriptor PCR should be set to the access line speed. These PVCs have zero equivalent bandwidth associated with them and will therefore be routed on the shortest path from origin to destination.

There is no policing or CAC (connection admission control) on best effort PVCs. Cells will be delivered to the destination as long as large queues do not cause UBR buffers to overflow.

The VBR-NRT service should use the PCR CLP=0+1, SCR CLP=0, MBS CLP=0, tagging traffic descriptor (see definition and example in "ATM terms explained"). This will provide the user with a traffic contract and a specified level of service. Non-conforming CLP=0 cells will be tagged as CLP=1 and passed through the network unless the rate of the cell stream exceeds PCR in which case the violating cells will be dropped.

Therefore, it is imperative that CPE equipment implement traffic shaping. If equipment does not implement effective traffic shaping then it is possible that large numbers of cells will be discarded at the ingress by UPC policing (see ATM terms explained).

The ATM traffic contract is more complex than the frame relay traffic contract because it clearly specifies peak cell rates (PCR) and maximum burst sizes (MBS) as well as sustained or average cell rates (SCR).

This is a benefit in that it provides users with more granularity in specifying their traffic characteristics. However, it also creates more complexity for users in determining traffic contracts and associated service pricing that is optimal for their needs.

2- Classical IP over ATM
Service description

Classical IP over ATM is defined in RFC 1577. The IP over ATM service offering is depicted in Figure 2.
IP over ATM
Figure 2

IP over ATM


The classical IP over ATM service can be offered by connecting routers to the ATM network switches using the ATM Forum UNI 3.1 interface. All the RFC 1577/1483 software for transporting IP over ATM resides in the CPE router.

The ATM network is divided into logical IP subnets (LIS). Each LIS contains an ATM ARP server that resolves the ATM address for an IP address. The service provider must provide a 1577 ARP server for each LIS.

In Figure 2, only one LIS is depicted. In a large ATM network multiple LlS's will be implemented. To execute this, a service provider needs at least one router operating as a 1577 ATM ARP server.

QoS class
It is recommended that the 1577 service use either the UBR or ABR QoS class. The major reason for using UBR is that today routers only support UBR. Another important reason for using UBR or ABR is because routers require the full bandwidth of the physical circuit to transmit data. This means that for a VC:

PCR=SCR=Line Speed

If UBR/ABR is not used then CAC will reserve an inordinate amount of bandwidth for an RFC 1577 VC.

3- LAN Emulation (LANE)
Service description
A LANE service is an alternative to a classical IP over ATM service for providing LAN interconnectivity across a public ATM network. LANE service is very similar to spanning tree bridges. Packets are forwarded based on MAC addresses.

LANE is a good method of interconnecting existing routers, LAN switches, and ATM hosts across an ATM infrastructure. Since LANE operates at the MAC layer it supports multiprotocol traffic at layer 3. LANE also supports MAC multicast and other characteristics of an Ethernet.

The configuration of a LANE service is depicted in Figure 3.
LANE services
Figure 3
LANE services


User devices are configured as LAN emulation clients (LEC). LECs can be routers, LAN switches, or ATM hosts running LANE software and connecting to the ATM network with the UNI 3.1 interface.

The ATM network can provide the service with a combination of ATM switches and LANE servers. The LANE servers connect to the ATM network cross a UNI 3.1 interface.

The following types of servers are needed for LANE:

LAN Emulation Server (LES): The LES implements the control function for a particular emulated LAN (ELAN). There is only one logical LES per ELAN, and to belong to a particular ELAN means to have a control relationship with the ELAN's LES. Each LES is identified with a unique ATM address.

Broadcast and Unknown Server (BUS): The BUS is a multicast server that is used to flood unknown destination address traffic and forward multicast and broadcast traffic to clients within a particular ELAN. Each LEC is associated with only a single BUS per ELAN, but there may be multiple BUS's within an ELAN.

LAN Emulation Configuration Server (LECS): The LECS is an entity that assigns individual LANE clients to particular ELANs by directing them to the LES that corresponds to the ELAN. There is one LECS per administrative domain that serves all ELANs within that domain.

Figure 3, represents a single ELAN with single servers. It is also possible to have multiple ELANs within the same physical ATM network.

QoS class
LANE clients only support UBR. Also another important reason for using UBR or ABR is because LAN switches and routers require the full bandwidth of the physical circuit to transmit data. This means that for a VC:

PCR=SCR=Line Speed

If UBR/ABR is not used then CAC will reserve an inordinate amount of bandwidth for a LANE VC.

Multipoint circuits ATM multicast circuits will be set up using an SVC from the BUS to all LECs. This multicast circuit is used for broadcasting unknown MAC addresses. All other circuits in the LANE ATM network are unicast circuits.

4- Voice over ATM
Service description
Currently, most voice is carried over ATM as CBR traffic. While this is a simple way to carry voice, it is not an efficient use of the ATM network.
As voice over ATM technology progresses, voice CPE equipment will incorporate native ATM interfaces with VBR service contracts and traffic shaping capabilities.
Some of the options for offering voice over ATM service are depicted in Figure 4. Voice carrying devices that could be connected by public ATM networks are:

Voice over ATM.jpg
Figure 4
Voice over ATM


QoS class
The VBR-RT QoS class should be used for voice over ATM VCs.

Traffic contract
If VBR-RT services are used to carry voice (as opposed to CBR), then SVCs (switched virtual circuit) could be used to set up and break down calls. In this case the traffic contract will be signalled by the SVC. Because ATM voice technology is new and also because the performance of voice compression is rapidly improving, it is difficult to know exactly what traffic parameters to use. With time and experience it is possible to have good estimates of traffic parameters to use when connecting voice to an ATM network. Some traffic parameter examples for voice are given below, these are not meant to be definitive but serve as a good starting point to evaluate a particular service.

SVC/PVC
Standard telephony service should use SVCs. Broadcast voice could use multipoint PVCs.

Multipoint circuits
Only broadcast voice could use multipoint PVCs.

5- Videoconferencing
Service description
Currently, most video is carried over ATM as CBR traffic. While this is a simple way to carry video, it is not an efficient use of the ATM network. As video over ATM technology progresses, video CPE equipment will incorporate native ATM interfaces with VBR service contracts and traffic shaping capabilities. Videoconferencing is a service that can be offered on an ATM backbone. Some of the possible configurations for Videoconferencing are depicted in Figure 5.

Video conferencing.jpg
Figure 5
Video conferencing


PCR SCR MBS
300 150 32
600 300 32
1000 500 32
1500 750 32
2000 1000 32
Table 3.
Video conferencing over ATM traffic parameters example


PCR SCR MBS
25 20 15
25 20 30
50 25 15
50 25 30
100 50 15
100 50 30
100 75 15
100 75 30
Table 4.
Video broadcast over ATM traffic parameters example


Videoconferencing equipment can be separated into two main categories room equipment and desktop equipment.

Room equipment could have direct ATM interfaces (UNI 3.1) or interface to an access concentrator on an El circuit. If the equipment interfaces to an access concentrator, the video stream is converted to ATM at the concentrator and the concentrator interfaces to the ATM network via UNI 3.1. Desktop video typically runs across a LAN. In this case the interface to the network is via a router or LAN switch with an ATM connection. The ATM video conference service offering should allow interfaces to all types of videoconferencing CPE.

QoS class
If CPE equipment supports VBR with traffic shaping, then the VBR-RT QoS class should be used for videoconferencing. Otherwise the CBR QoS class must be used.

Traffic contract
If VBR-RT services are used to carry video (as opposed to CBR), then SVCs could be used to set up and break down calls. In this case the traffic contract will be signalled by the SVC. Because ATM video technology is new and also because the performance of video compression is rapidly improving, it is difficult to know exactly what traffic parameters will be used. However, with time and experience it is possible to have good estimates of traffic parameters to use when connecting voice to an ATM network. Some traffic parameter examples for videoconferencing are given below, these are not meant to be definitive but serve as a good starting point to evaluate a particular service.

SVC/PVC
SVCs are preferable for videoconferencing services.

Multipoint circuits
Videoconferencing should only use point to point VCs.

6- Video broadcast
Service description
A video broadcast service offering could be implemented on an ATM network using a variety of access technologies. Examples of video broadcast access are depicted in Figure 6. In video broadcast applications the video is broadcast by a single source and received at multiple stations. Both the source and the receivers could be connected to the ATM network via a variety of methods including: direct UNI 3.1 connections; connections through ATM access concentrators; connections through LANs.

Video broadcast.jpg
Figure 6
Video broadcast


QoS class
If CPE equipment supports VBR with traffic shaping, then video broadcast should use the VBR-RT QoS class. Otherwise the CBR QoS class must be used.

Traffic contract
If VBR-RT services are used to carry video broadcast (as opposed to CBR), then either SVCs or PVCs could be used. SVCs would generally be used for video on demand and multicast PVCs would be used for broadcast video. In the case of SVCs the traffic contract will be signalled by the SVC. Because ATM video technolog, is new and also because the performance of video compression is rapidly improving, it is difficult to know exactly what traffic parameters will be used.

However, with time and experience it is possible to estimate traffic parameters for Video Broadcast applications. Some traffic parameter examples are given below, these are not meant to be definitive but serve as a good starting point to evaluate a particular service.

SVC/PVC
Video broadcast could use either multicast PVCs for broadcast or SVCs for video on demand.

Multipoint circuits
All VCs are multicast.

PCR SCR MBS
2000 1000 32
4000 2000 32
6000 3000 32
8000 4000 32
10000 5000 32
Table 5.
Video broadcast over ATM traffic parameters example


7- Circuit emulation
Service description
Circuit emulation services could be provided by a combination of an ATM switched network and ATM access multiplexers. A circuit emulation network is depicted in Figure 7. The circuit emulation service provides El, or higher speed circuits for legacy TDM equipment. The circuits are directly terminated into an ATM access concentrator and then the CBR traffic is carried across the ATM network.

QoS class
The CBR QoS class should be used for circuit emulation.

Circuit priority
N/A. Circuits priority is only used for VBR circuits.

Traffic contract
The PCR CLP=0+1 traffic descriptor should be used for the circuit emulation service. The value of PCR should be equivalent to the line speed of the circuit being emulated.

SVC/PVC
All circuits are PVCs.

Multipoint circuits
All circuits are unicast.

ATM cct emulation.jpg
Figure 7
ATM circuit emulation


ATM terms explained
ATM traffic contract
Is a set of parameters that defines an agreed traffic profile between a device and the ATM network. Three parameters are defined:
  1. PCR = Peak Cell rate
  2. SCR = Sustained Cell Rate
  3. MBS = Maximum Burst Rate


The traffic contract for a connection is described by a traffic descriptor.

ATM frame descriptor
Characterises the ATM virtual connection in terms of peak cell rate (PCR), sustainable cell rate (SCR), and maximum burst size (MBS). This information is given to a connection by the network and is used to allocate resources (eg, bandwidth, buffering) in the ATM network. In general, the ATM traffic descriptor for supporting multiprotocol interconnection over ATM will be driven by factors such as the capacity of the network, conformance definition supported by the network, performance of the ATM endsystem and (for public networks) cost of services.

Example of traffic descriptor:
PCR CLP=0+1, SCR CLP=9,
MBS CLP=0, tagging


In this example the PCR is a rate (say 100 cell/sec) for aggregate traffic with CLP bit set to zero and for traffic with CLP=1 (discard eligible) the SCR is also a rate (say 60 cell/sec) for CLP=0 traffic and the MBS is a number (say 200 cells) for CLP=0. Any traffic that does not conform to these parameters will have its CLP bit set to 1 (tagging) so that it becomes discard eligible or dropped if it exceeds PCR or violates the traffic descriptor in any way.

The CLP (cell loss priority) bit is part of the ATM cell header see below:

payload.jpg
Figure 8
Payload, headers and identifiers


Traffic shaping
The use of buffers to limit surges that can congest a network. Data is buffered and then sent into the network in regulated amounts to ensure that the traffic will fit within the traffic contract parameters for a particular connection.

Peak cell rate (PCR)
PCR is the maximum allowed traffic to a connection in units of cell/seconds.

Sustainable cell rate (SCR)
SCR is the long term average cell rate in cell/seconds that a source transmits.

Maximum burst size (MPS)
MBS is the largest number of cells that can be temporarily transmitted at a rate higher than the PCR.

Usage parameter control (UPC) policing
Is a process that runs in an ATM network to measure actual traffic across a given connection and compares it to the traffic descriptor for that connection. Cells violating the descriptor parameters can be tagged (ie CLP bit set to 1) or be discarded if congestion develops.

Click here to go back to my Home Page

Enter your email address to receive email when this page is updated.

Your Internet email address:


This page has been accessed times.

Last revised: Sunday, 21 June 1998