The years since the World Wide Web was first developed have seen an explosive growth in the use of telecommunications and data communications services. Coupled with this growth in the market for communications technology has been a major technological push, which is leading to a burgeoning in the number of services available, a trend which is likely to persist for the foreseeable future.

Consequently the period since roughly 1990 has seen major change in the communications industry. The major factors promoting change have been the following:

  • The ending of monopolies in the provision of telecommunications services
  • The revolution in network capacity precipitated by the introduction of fibre to the network
  • The explosive growth of the Internet
  • The increasing levels of market penetration for IT products with datacomms capabilities (smartphones, etc.)
  • The convergence of the telecommunications, data communications, and cable television industries
  • The widespread availability of and interest in Video-on-Demand Services.

These developments pose considerable dilemmas for network service providers. The bandwidth requirements of users are mushrooming, the range of services is escalating, and the number of users is skyrocketing. How can a network be designed and maintained to meet the requirements of the market?

The response has been to develop a range of new protocols (ATM, then MPLS, then GMPLS, etc.) and to augment existing protocols such as IP (with Diffserv, Intserv, RSVP, SBM).

The goal in deploying these and even newer technologies is to keep costs down, whilst increasing the number of customers and the bandwidth available to each, all while maintaining customer satisfaction. A network which meets these conflicting requirements may be termed a broadband network.

This module only considers wired networks (both coppper and optical fibre) - the special techniques needed to get high performance in wireless networks are the topics of other modules.

The key concerns facing the designer of a broadband network are:

  • The network should be scaleable
  • The network should provide Quality of Service

We will explore what these terms mean, and how these conflicting requireents can be met.

The ideal protocol

Why don’t we focus on the Internet Protocol (IP)? It’s the most widely used protocol, right? Well, there are three major reasons:

  • IP is a relatively old protocol, and as protocols age, they get complicated. Newer protocols tend to be simpler protocols (although this is less true after they are around a while, as mission creep sets in).
  • IP does not support Quality of Service (QoS) without relatively complicated extensions like RSVP and DiffServ (look them up) - newer protocols such as ATM and MPLS (nominally) do. (Since the QoS extensions to IP are optional, in practice this means they are often unavailable.)
  • Many of you will already have had exposure to IP.

If we were to design the Internet from scratch to support the demands currently based upon it, we would choose rather different protocols to run upon it. IP is (fairly) scaleable but natively offers no support for QoS. I could discuss a hypothetical ideal protocol which meets such requirements, but the resulting discussion would be very dry.

Instead we will pick a reference protocol whose merits are the reverse of those of IP (it supports QoS but is not as scaleable as IP). I use ATM as the reference protocol for this module. It’s not an ideal protocol, by any means (its chief limitation being the fixed packet length), but it contains many of the features a modern network protocol should possess. We could instead discuss Multiprotocol Label Switching (MPLS), which is more IP-friendly than ATM, but the requirement for IP compatibility makes the QoS features of MPLS harder to follow. Most of what you’ll learn will be valid for any QoS-friendly protocol, although the names may change (one person’s route pinning is another person’s explicit routing is another person’s virtual circuits). I will endeavour to provide details of the nearest equivalent feature of IP, when appropriate.