. 91
( 132 .)


Rest of
HEC Data

4 1 53

Figure 7.13 Header checking

7.5.3 Inverse multiplexing for ATM (IMA)

IMA is a standard de¬ned for ATM which allows an ATM cell-stream to be inverse
multiplexed and carried by multiple physical links, with the original stream retrieved
at the other end of the links. It de¬nes a new sublayer at the physical layer, the IMA
sublayer. It may be used, for example, to transport an ATM stream over several E1/T1
lines, grouped to provide a suitable rate for the ATM, effectively creating a higher-
bandwidth logical link, referred to as an IMA group. The inverse multiplexing process is
performed on a cell-by-cell basis in a round-robin scheme. The concept is illustrated in
Figure 7.14 below.
Also de¬ned is a new type of operation, administration and maintenance (OAM) cell,
which can either be an IMA control protocol (ICP) cell or a ¬ller cell. The ICP cell
contains information on the format in which the cells are being sent on the physi-
cal link, referred to as an IMA frame. This allows the receiver to adjust for differing
delays on the physical links, since the frame format is known and timings can be estab-
lished from the arrival times of ICP cells. The ¬ller cells are used when there are no
ATM cells to send. They are inserted to maintain timings and decouple the cell rate
from that of the physical links. The receiver will remove ¬ller cells when reconstructing
the stream.
In summary, Figure 7.15, shows the physical layer with the various mechanisms dis-
cussed in their relative positions.

IMA group IMA group


ATM cell stream
ATM cell stream PHY PHY


Figure 7.14 Inverse multiplexing for ATM

ATM Layer

TC Sublayer

PMD Sublayer

Figure 7.15 The physical layer


The ATM layer provides a connection-oriented service and it is at this layer that the virtual
circuit and path are established. However, ATM differs from most connection-oriented
protocols in one major respect, as no acknowledgement is given for cells received, but
it does guarantee that cells arrive in order provided that they travel on the same virtual
circuit. Nevertheless, cells may be discarded should the network become congested.
At the ATM layer, two boundaries are considered, as shown in Figure 7.16. The ¬rst
is that between a host and an ATM network, referred to as the user to network interface
(UNI), and the second between two switches on the network, the network to network
interface (NNI). The NNI de¬nes the relationship between the two switches, regardless of
whether they are on the same or different networks, or indeed lie on a boundary between
a public and private ATM network. However, there are different protocols de¬ned to deal
with these different NNI relationships.
It has been seen that the ATM cell structure consists of 5 bytes of header and 48 bytes
of data. The format of the header is slightly different for the UNI and the NNI, as shown
in Figure 7.17.
As can be seen, the difference is in the generic ¬‚ow control (GFC) ¬eld which is
unused and overwritten once it reaches the ¬rst ATM switch with an extended virtual
path identi¬er (VPI), since one switch can (and generally will) be connected to many
hosts. The GFC is intended primarily for proprietary use between a user and a switch,
allowing for the ¬rst switch to control the ¬‚ow of information from a host. As an example,
a switch may use the GFC to halt a host™s cell ¬‚ow. However, virtually all manufacturers
ignore it and for UNI cells, the GFC ¬eld is ¬lled with 0000. Note that it has no relevance
to ¬‚ow control within the network.
The VPI speci¬es the virtual path that should be used. For ef¬cient routing arrange-
ments, ATM switches may only store information about the path and not the individual
circuits, meaning that a number of virtual circuits take up only one entry in the translation
table of the switch.

ATM Network


Switch Switch

Figure 7.16 ATM interfaces

UNI Cell NNI Cell
8 1 8 1
Generic Flow Control Virtual Path Identifier Virtual Path Identifier
1 1

Virtual Path Identifier Virtual Channel Identifier Virtual Path Identifier Virtual Channel Identifier
2 2

Virtual Channel Identifier Virtual Channel Identifier
3 3

Virtual Channel Identifier Payload Type Virtual Channel Identifier Payload Type
4 4

Header Error Control Header Error Control
5 5

Data Data
48 byte payload 48 byte payload
53 53

Figure 7.17 UNI and NNI headers

The 16-bit virtual circuit identi¬er (VCI) identi¬es a particular virtual circuit within a
given virtual path. Since the user can specify an 8-bit VPI and a 16-bit VCI, this allows
for the allocation of 256 paths, each containing up to 64 k circuits. In practice, the ¬rst
31 circuits are reserved for special purposes such as signalling.
Idle cells, which are inserted to maintain timing when synchronous communication is
required, have the format shown in Figure 7.18.
The 3-bit payload type (PTI) ¬eld identi¬es the type of cell being transmitted. The most
signi¬cant bit (msb) indicates whether it is a data cell (PTI msb = 0) or a management
information cell (PTI msb = 1). The middle bit is used as a congestion indicator, when the
msb is 0 (i.e. for data cells). If congestion is experienced, this bit is set to 1 by the switch
so that the destination host is provided with an indication that there is congestion on the
network. The destination can then feed this information back to the source, otherwise the
source would be oblivious as to whether it is contributing to network congestion. In con-
trast to ATM, the TCP/IP protocol provides an acknowledgement of all data transmitted.
The source can then use this acknowledgement mechanism to determine if the network
is congested, and deal with this congestion using a sliding window implementation.
Finally, the least signi¬cant bit (lsb) of a data cell is used to indicate a type 0 or 1 cell.
This is used by one of the adaptation layers, AAL5, and is described shortly. A summary
of the values for the PTI ¬eld is presented in Table 7.7.
The cell loss priority (CLP) bit is used to identify the priority of a cell. Should the
network become congested, the switches will ¬rst drop lower priority cells, CLP = 1.
The HEC checksum is as explained in the previous section.


0000 0000 0000 0000 0000 0000 0000 0001 HEC payload

byte 1 byte 2 byte 3 byte 4 byte 5 bytes 1-48

Figure 7.18 Idle cell format

Table 7.7 Payload type values
PTI Signi¬cance
000 User data cell, no congestion, cell type 0
001 User data cell, no congestion, cell type 1
010 User data cell, congestion, cell type 0
011 User data cell, congestion, cell type 1
100 Maintenance between adjacent switches
101 Maintenance between source and destination switches
110 Resource management cell
111 Reserved

The payload section contains the data. However, note that all 48 bytes need not be
data, since the higher layers will utilize some of this space to transfer their respective
header information.


The AAL was conceived to provide a good interface between the different kinds of
applications (voice, video and data) and the ATM network. Since the lower layers do not
provide control functions, such as error and ¬‚ow control, which are generally needed by
applications, this layer is designed to bridge the gap. To de¬ne a suitable interface, the
ITU-T considered that applications fell into three broad groupings as in Table 7.8.
This gives eight categories of service, of which four were chosen to support; the other
four were considered not to be useful, or make logical sense. The four protocols de¬ned
for the services were termed AAL1 to AAL4, as shown in Table 7.9.
However, it was decided that since there was a lot of similarity between AAL3 and
AAL4, they would be combined into one protocol, AAL3/4. Unfortunately, the computer

Table 7.8 Service categories
1 Real-time Non-real-time
2 Constant bit rate Variable bit rate
3 Connection-oriented Connectionless

Table 7.9 AAL service protocols
Mode Connection-oriented Connectionless
Bit rate
Constant Variable Constant Variable
Timing Real- Non- Real- Non- Real- Non- Real- Non-
real- real- real- real-
time time time time
Protocol AAL1 AAL2 AAL3 AAL4

Table 7.10 Revised AAL protocols
AAL protocol Traf¬c Example UMTS application
1 Constant bit rate E1/T1 voice circuit emulation
2 Real-time variable bit rate traf¬c Voice CODECS Packet video, data
3/4 Variable bit rate Best-effort data Unused
5 Data IP traf¬c to core network

Service Specific Part
sublayer CS AAL
Common Part

Segmentation and reassembly sublayer

Figure 7.19 AAL sublayers

industry felt that none of the protocols was suited to their data transfer needs and so
an extra protocol, AAL5, was implemented. The revised protocols are summarized in
Table 7.10, with an introduction to their use within UMTS.
The AAL is split into two sublayers: the convergence sublayer (CS) and the segmen-
tation and reassembly (SAR) sublayer, as shown in Figure 7.19.
The CS provides the interface to the applications for a particular adaptation layer. The
lower part of the sublayer is common to all applications while the upper service-speci¬c
part is application speci¬c. The CS accepts messages from applications and splits them up
into units for transmission. A unit ranges from 44 to 48 bytes payload, and is dependent
on the AAL used since some of the AAL protocols will add their own header information
at the CS. When the CS receives cells, its role is to take the cells from the SAR sublayer
and reconstruct the original message or data stream.
The SAR sublayer takes the cells and passes them to the ATM layer for transmission.
Again, depending on the protocol, it may also add its own header information. As its


. 91
( 132 .)