Side_1_360

work were the access link is in use during a

phone conversation, but idle most of the time.

Installing a capacity much higher than the aver-

age load is a reason for the rather high cost of

the access network.

As many types of applications and users will be

connected, a range of access capacities could be

requested. Commonly, however, rather fixed

capacities are used, e.g. offering Asymmetric

Digital Subscriber Line (ADSL) downstream

rates as 384, 768 and 1024 kbit/s. This also con-

tributes to the potential capacity of the access

network usually being poorly utilised.

A core domain usually carries traffic from

a greater set of customers resulting in higher

capacity links and routers. A certain averaging

effect of the traffic, e.g. during the day is also

commonly observed, for example related to the

different customer types and use of services.

The edge of the core network usually has a num-

ber of features related to individual users, like

access lists and monitoring mechanisms. Edge

routers will then implement such features.

Towards other operators, border routers, possi-

bly with similar functionality can be present.

Providing efficient traffic handling implies that

appropriate solutions must be available in all

the domains involved. However, for customer

equipment, it is usually expected that the cost of

providing capacity is relatively low, implying

that capacity is added in case a performance

problem emerges. For other domains mecha-

nisms for differentiated handling of services

(and traffic flows) seem to be gradually intro-

duced. In addition to differentiating between ser-

vices, differentiating between customers would

also be likely (although this could again be seen

as a kind of service differentiation).

When looking for potential performance bottle-

necks it is essential to include the end systems

as well as the processing capabilities in all

domains. This means that efficiency of protocol

software (as well as software of other traffic

handling mechanisms) is a pivotal point. As the

capacity evolution of transmission outpaces the

computer capacity growth, the system designers

may rather concentrate on achieving high trans-

fer speed (putting data on the output interface)

than optimised utilisation of the transmission

bandwidth when seen from an end system point

of view. The network operator view may be

more balanced, however, as a huge number of

traffic flows will be present.

2.3 Layering

Similar to the OSI model (Open System Inter-

connection), the Internet-related protocols can

also be depicted as a hierarchy, see Figure 6.

Numerous protocols could be used below IP,

like Asynchronous Transfer Mode (ATM) and

Point-to-Point Protocol (PPP). A number of pro-

tocols could also be used in the transport layer,

although Transmission Control Protocol (TCP)

and User Datagram Protocol (UDP) are amongst

the most popular ones.

Figure 5 Introducing
equipment for concentration
and multiplexer to reduce the
cost for dedicated access lines
per user

Figure 6 Layers and example
of protocols related to IP

ADSL

DSLAM

edge

router

= concentrator/multiplexer

layers

IP

ICMP

ATM

PPP

TCP UDP

Telnet FTP HTTP Routing

protocols

Voice/video

codec