ATM quality of service and its use in IP networks

Contents of the paper

Glossary
References

Further Information
 

1 Introduction

The ATM Forum is developing Asynchronous Transfer Mode networking which provides real-time networking support. Also the Internet Engineering Task Force (IETF) is currently developing an integrated service model which is designed to support real-time services on the Internet.
The use of ATM in the Internet as a link layer protocol is already occurring, and both the IETF and the ATM Forum are producing specifications for IP over ATM [6].

2 Introduction to ATM

This support for guaranteed QoS connections is achieved when the ATM switches communicate their QoS needs and traffic characteristics to the network, and the network communicates admission control decisions back to them. Once the desired resources have been reserved they are kept for the lifetime of the connection. This mechanism (Resource Reservation) is explained in detail in chapters 2.2 and 2.6.

ATM Interfaces

The ATM network consists of ATM switches interconnected by point-to-point links or interfaces, these interfaces can be of two types:

1. User-Network Interface- UNI
2. Network-Node Interface- NNI

Figure 1: ATM Network Interfaces [3]
 

2.1 ATM Layers

ATM Adaptation Layers

The AAL layers provide an organized method of supporting multiple services over a common network infrastructure [2].

These layers translate the information received from the user (voice, video, frame relay, X.25, ...) and the ATM cell provided by the ATM layer.
They assure the appropriate service characteristics and divide all types of data into the 48 octet payload that will make up the ATM cell.
The data from the different sources is handled in unique way depending on the source, but when the data is transmitted across the network it is treated as the same kind of data. Only inside the AAL the information about the different interfaces is used, later in the ATM layer this information is ignored.

There are five types of AALs (AAL 1-5) depending on the type of service that the AAL is dealing with. See table 1 in chapter 2.5 for a correspondence between the AAL layers and the types of services.

ATM Layer

This layer adds the cell header to the 48 octet payload received from the AAL.

Its main functions are:

• Usage of  Virtual Channel Identifiers (VCI) and Virtual Path Identifiers (VPI).
• Multiplexing and Switching of cells.
• Providing virtual connections between end points
• Maintaining the QoS (Traffic contract at call set-up and Policing during the connection)

• GFC Generic Flow Control
• VPI Virtual Path identifier: it allows for virtual path switching in the network.
• VCI Virtual Channel Identifier: it allows for virtual channel switching in the network.
• PTI Payload Type Identifier: to differentiate between users and management data.
• CLP Cell Loss Priority: it indicates whether the cell can be discarded if necessary. This bit can be set by the ATM terminal if the cost of the service is reduced for low priority traffic, to set priorities for different types of traffic or if the traffic contract has been violated.
• HEC Header Error Control: it allows the system to execute multiple error checking in the Physical layer. If the cell fails the error check, it is instantly discarded by the network and no further action is taken [2].

2.2 ATM Resource Reservation, Switching and Routing

Cells belonging to a particular service or destination are assigned the same "Virtual Channel" in the AAL layer, and many virtual channels share a single link and placed together into "Virtual Paths" in the AAL layer.

The virtual paths are used for QoS management in the following way:

• A VP may be given some guaranteed QoS
• Their bandwidth can vary as the VC bandwidth varies [1]. This way bandwidth can be reserved for some services and be limited for others.
• Allocation of delay priorities, by switching the VP's to different queue's.

There are two types of ATM switches: VC switch (it switches VC's from one VP to another) and VP switch (it switches VP's and does not take into account the VC's inside).

The functionality of an ATM switch related to QoS support can be resumed as follows:

1. Connection set-up

Before sending any data a connection must be set-up along the network (from the source to the destination). The switch wishing to set up the connection sends signaling packets with VPI=0 and VCI=5.  These packets are routed from switch to switch until they arrive to their destination. Once the connection has been accepted, at the destination system, the user data will use the same path.

Figure 2: Connection Setup through ATM Signaling (SVC) [3]

When the switch receives a connection request from another switch, it performs the Connection Admission Control function (CAC). The switch checks the traffic parameters and QoS requirements for the connection, and determines whether setting up the connection violates the QoS guarantees of already established connections.  If not, then it accepts the connection.

The VC routing protocol must ensure that a connection request is routed along a path which leads to the destination and also has probability to be accepted in the forthcoming ATM switches.

Figure 3: Connection Admission Control [3]

Once the connection is established, the switch receives the cells with their VCI and VPI. It checks these values against a table, and finds out the new values of the VCI and VPI, and towards which output-port it should retransmit the cell.

2.3   ATM QoS parameters

Applications must specify the QoS they require from the network and the type of traffic that they wish to submit.
In ATM, the Private Network-to-Network Interface (PNNI) protocol communicates QoS information along with routing information, and the network nodes can utilize this information to establish paths for the required QoS.

The QoS Parameters for ATM are divided in additive (their values are accumulated through the network) and  non-aditive (it is a configured link and node attribute).

Additive parameters:

• Cell Delay Variation (CDV): cells belonging to the same VC which suffer different delays due to varying queue lengths in the ATM switches. Delays are caused by Cell packetisation delay, Queuing delay in switches, De-packetisation delay, Standard switching delays and propagation delays across the network.
• Maximum Cell Transfer Delay (Max CTD)

• Cell Loss Ratio (CLR): it may be caused by transmission errors that corrupt cell headers (and therefore are discarded), congestion due to traffic peaks (if the CLP is set to 1, and the cell has been queuing too long they can be discarded) or equipment failure.

2.4 ATM traffic parameters

•      Peak Cell Rate (PCR)
•      Cell Delay Variation Tolerance (CDVT)
•      Sustainable Cell Rate (SCR)
•      Burst Tolerance (BT)
•      Minimum Cell Rate (MCR), for ABR only.

The traffic characterization provided by an application or user is used by the network to make decisions about how to provide the desired quality of service to this application. It is also used to assess the effect the new flow will have on the service provided to existing flows.

2.5 ATM Service categories

Table 1: Service Category Attributes and Guarantees
 

Service Category Attributes and Guarantees
Service Category *	Traffic Description	Correspondent AAL layer	Min Loss (CLR)	Delay/Variance	Bandwidth	Typical Use
Constant Bit Rate (CBR)  [Deterministic Bit Rate (DBR)]	PCR	1	x	x	x	Real time, QoS guarantees. Video-conferencing, telephony, distance learning, pay per view TV, audio library.
Real-Time Variable Bit Rate (rt-VBR)	PCR, SCR, MBS	2	x	x	x	Statistical mux, real time. Voice with silent suppression (cells are transmitted only when the person is speaking).
Non-Real-Time Variable Bit Rate (nrt-VBR)  [Statistical Bit Rate (SBR)]	PCR, SCR, MBS	3/4	x	no	x	Statistical mux. Data transfer such as Banking transactions.
Available Bit Rate (ABR)	CR, MCR+ behavior parameters	3/4	x	no	x	Resource exploitation, feedback control. LAN interconnection, critical data transfer such as banking applications.
Unspecified Bit Rate (UBR)	(PCR)	5	no	no	no	Best effort, no guarantees. Text and Image transfer, messaging. Connection-less upper layers protocols data such as IP.

* ATM forum "Service category" [ITU-T I.371 "ATM Transfer Capability"]

CBR: for constant (maximum) bandwidth services. These kind of connections need a fixed amount of bandwidth available for the whole duration of the connection. They might transmit at Peak Cell Rate (PCR) for sometime and later remain silent. Typically these connections receive high priority to minimize the amount of latency and jitter experienced by the cells in these connections [3].
rt-VBR: for time sensitive applications, its transmission rate varies with the time and delay is not tolerated.

nrt-VBR: for applications with bursty traffic and tolerating with delay and delay variations.

ABR: for applications capable of varying its emission rate. Those applications require a low cell loss ratio, but do not have tight requirements for throughput and delay. The ABR benefits from the "elastic bandwidth allocation", an ATM property that allows the sources to increase/reduce the information rate depending on the network availability.

UBR: applications which tolerate delays and which do not have special QoS requirements. These applications are expected to transmit non continuous bursts of cells. This service class is the closest to the best effort service of traditional IP.

2.6 QoS support: Traffic contract and Policing

1. The characteristics of the traffic that the user will send.
2. The QoS that the network should provide

The policing function estimates the parameters of the incoming traffic and takes action if the measured traffic exceeds the agreed parameters.
The ATM switch performs this checking with an algorithm, the Generic Cell Rate Algorithm (GCRA) or "continuous leaky bucket".

The Generic Cell Rate Algorithm

The GCRA provides a rule to measure whether cell streams accomplish with their bandwidth parameters as specified in the traffic contract.
The Increment parameter, I, is the fill rate of the bucket. It is related to the distance in time between two consecutive cells of the same virtual connection.
The Limit parameter, L, represents the maximum value of the bucket fill level.

As a stream of cells arrive at the network interface, the contents of the bucket is increased by I for each cell that arrives. If the cells arrive at a faster rate than the agreed the bucket becomes full. The capacity of the bucket is L+I and a cell that causes the bucket to overflow is marked as non-conforming [2].
Such cells will be either discarded or the Cell Loss Priority bit will be set to 1.

3 IP QoS vs ATM QoS

3.1 IP as a best-effort datagram transmission

3.2 IP service classes

Nowadays the method of QoS specification for the Internet is to specify a `service class' and some set of parameters, depending on the service class.
The currently proposed service classes are:

•  guaranteed, which provides a mathematically guaranteed delay bound
•  predictive delay, which provides a probabilistic delay bound
•  controlled delay, which merely tries to provide several levels of delay which applications may choose between.
•  and the existing `best-effort' class.

3.3 Differences between both QoS characterizations

In the Internet community, it is assumed that traffic will in general be bursty and that bursty traffic can be characterized by a `token bucket'.

ATM in turn, does not expect all traffic to be bursty. Furthermore, ATM in some classes also requires specification of peak cell rate, whereas peak rates are not currently included in the IP traffic characterizations.
One of the functions that must be performed in order to carry IP traffic over an ATM network is therefore a mapping from the characterization of the traffic as supplied to IP to a characterization that is acceptable for ATM. [6]

QoS characterization

Since the goal is to carry IP efficiently over ATM networks, it is necessary to establish mechanisms by which QoS specifications for IP traffic can be translated into QoS specifications that are meaningful for an ATM network. [6]

An ATM virtual circuit is established based upon a user's traffic characteristics and network performance objectives. These characteristics which include delay and throughput requirements can only be defined by the application level (at the transport level or above) as opposed to the internetworking level.
For instance, a file transfer application transferring a 100 MB file has very different link level performance requirements than a network time application.
The former requires a high throughput and low error rate connection whereas the latter could perhaps be adequately serviced utilizing a best-effort service.
Current IP does not provide much support for a quality of service specification and provides no support for the specification of link level performance needs by an application directly. This is due to the fact that only a single type of link level performance is available with link technologies like Ethernet.  As a result, all applications over IP today receive the same level of link service [9].

Routing

ATM computes a path at connection set-up and leaves the path in place until the connection is terminated or there is a failure in the path.
An ATM cell only carries information identifying the connection and no information about the actual source and destination of the cell.
An ATM switch checks a table of the established connections (VP/VC) that map to the next hop device, without checking the final destination only the VP/VC values from the header of the cell.

However, routing decisions in IP are based on the destination address contained in every packet.
The IP router checks a table that contains the routes to all possible destinations and the routing decision is made based on the final destination of the packet.
If an IP path has been selected for a given QoS, changes in the route may mean a change in the QoS of the path.

4 Providing QoS in IP over ATM

 4.1 Real time services in an IP-over-ATM network

Real time applications can operate to some extent using best-effort delivery, but trading packet delay for correct reception is not always acceptable. Operating in the traditional mode for these applications results in reduced quality of the received information and, inefficient use of bandwidth.

Nowadays the IETF is developing a real-time service environment in which multiple classes of service are offered. This new environment will extend the existing best-effort service model to meet the needs of multimedia applications with real-time constraints [6] or Integrated Services Internet.
The Integrated Services Internet enhancements include traffic management mechanisms that closely match the traffic management mechanisms of ATM. For instance, protocols such as the Resource Reservation Protocol (RSVP) are being defined to allow for resource reservation across an IP network, just like ATM signaling allows this within ATM networks.

IP packets need not explicitly contain information parameters describing an application's traffic characteristics and network performance objectives (e.g., delay = low, throughput = 10 Mb/s). This information could potentially be mapped from resource reservation protocols that operate at the IP level [9].

Resource Reservation techniques

Two protocols have been proposed for Resource Reservation in IP [6]:

• ST-II  (sender based, hard state, dynamic QoS)
• RSVP (receiver based, soft state, dynamic QoS)

RSVP is a control protocol, that will be used by applications within IP end systems (receiver based) to indicate to nodes transmitting to them the nature (such as bandwidth, jitter, maximum burstiness, and so on) of the packet streams that they wish to receive. Intermediate systems, will  interpret RSVP control packets in order to perform admission control (analogous to ATM CAC) and allocate the resources required to support the requested traffic flows.
RSVP nodes will maintain "soft state" about such traffic flows, and will perform packet level traffic shaping, scheduling, etc., like when ATM switches perform traffic Policing to provide the agreed QoS.
RSVP can hence be thought of as providing very much the same traffic contract specification functions with respect to packet level traffic flows that ATM UNI and NNI signaling play with respect to cell flows .[3]
 

Figure 4: Mapping of the Integrated Services Internet into ATM [3]

RSVP routes traffic flows along source rooted point-to-multipoint paths, a flow which can be seen a stream of IP packets can be identified with VP and VC in ATM. Also the multicast protocols as for example Protocol Independent Multicast (PIM) and its routing capabilities could be identified with VC routing protocols in ATM.

Finally, only mention that the IP Version 6 protocol, IPv6,  will incorporate full support for integrated services through the use of such mechanisms and the definition of protocols like RSVP.
 

5. Conclusion

This is particularly useful if we take into account the newcoming multimedia applications which combine voice, video, image, and data traffic.
Nowadays the network layer protocols do not provided the appropiate QoS to the above mentioned applications, which will surely require it.

Therefore this paper presents as a solution, to provide this QoS, to run the applications or transport protocols directly across ATM.
Internet Protocol is the only protocol for which work is being done to enhance it to run directly across ATM networks, the mapping of both protocols´ QoS will require some work but it is possible and it would be benefitial.
 

Glossary of abbreviations

References

Further Information