Lecture 15: Internet Topology and Structure

"Leased Lines" and Router Interconnections

• The Internet is composed of many networks and/or subnets interconnected by routers.

• Where the networks and/or subnets to be connected are not geographically adjacent, it is normal for a permanent point-to-point link to be used. The historical terminology for a full-time telecommunications interconnect of this kind is a "leased line"^[1]. The name derives from the ancient technology of renting a permanently connected "phone line" plus the necessary modems at each end.

• A point-to-point inter-router link is commonly a separate subnet.

Organisations which wish to establish full-time links to the Internet for their internal networks need to establish a router-to-router link to another router which has Internet access. This can be done "cooperatively", where one Internet-connected organisation allows another to establish a link to its network, but nowadays is most commonly done by dealing directly with an Internet Service Provider.

^[1] The traditional "leased line" provides a so-called basic carriage service. Such services are not, in general, very profitable for a telecommunications company.

Example: Early AARNet

In this architecture, AARNet retained ownership of all of the routers, and the involvement of the prime telecommunications provider (Telstra) was limited to providing basic carriage service. In this model, Internet access between universities in Australia followed, to some extent, the cooperative model.

This architecture has now been replaced by a model whereby an C&W Optus operates as an Internet Service Provider, connecting Regional Networks (eg, the VRN in Victoria) in each state to each other and to the Australian and International Internet.

USA Internet Structure

In 1995, this structure was privatised, with a variety of commercial organisations (National Service Providers (NSPs) taking over the backbone functionality. Interconnections between these organisations occur at a number of Network Access Points (NAPs). NSPs are said to peer with other NSPs at the NAPs, with complex financial arrangments in place to "carry each other's packets".

Permanent Internet Connections for Business

1. Choosing an ISP who offers permanent Internet access at a reasonable price in the geographical area where the business is located.

2. Choosing the basic-carriage technology (or "leased-line" technology) which they will use to facilitate the connection to the ISP.

3. Managing several other issues, such as the purchase of a suitable router (which could be an ordinary computer system running appropriate software), possibly organising a domain name (and delegation), IP addresses, setting up necessary servers and lots more.

Example: Bigpond Direct

Telstra Corporation's permanent Internet service is offered as its Big Pond Direct service. Pricing is based on a installation charge, a fixed monthly charge plus a data volume charge. In addition, the business must separately acquire an "Access Method", see below. On interesting observation on the Telstra (and C&W Optus) Internet (IP) services is that charging is based on volume of data received, unlike the normal USA practice of charging on "pipesize" alone. This has some interesting implications...

"Leased Line" Technologies

• Traditionally: a permanently connected (or "nailed up") 'phone line and two modems -- in Australia, the ancient DATEL service worked this way. Bit rates up to 72kbps were available. Note that the term "nailed up" has survived to this day to describe any permanently available data circuit. Many businesses use a modern-day equivalent of this technology by taking advantage of Australia's "untimed local call" charging scheme, using a dial-in modem but holding the connection permanently open.

• DDS was an early (and fiendishly expensive) digital "leased line" service in Australia, with bit rates between 2400 and 48kbps. The original AARNet was based on this service.

• ISDN-based services. For example, a Telstra ISDN service can dedicate one or more B Channels as a permanently available data link, see later.

• Frame Relay services.

• Asynchronous Transfer Mode services.

ISDN Primary Rate

The NT2 (typically a PABX) connects to a special OnRamp NT1, at a point called the T interface. An S interface can be provided on the customer side of the NT2.

Note that in the USA and Japan, the primary rate service is instead 20B+D over a 1.544 Mbps ( called a T1) bearer.

Where all of the available B channels are dedicated as a "leased line" point-to-point data service, users refer to an E1 (2Mbps) or T1 connection speed.

ISDN-based "Leased Line" services

In its more recent OnRamp products, Telstra currently offers its OnRamp Xpress service. This is similar to a SPC, but charging is significantly more complicated! OnRamp Xpress charging is based on the idea of a monthly capped charge. Example:

A business purchases an OnRamp Xpress service to connect permanently to an ISP, or perhaps to another local business:

On Ramp 2	$50 per month
Xpress 1 Plan Fee	$90 per month
Xpress 1 limited cap	$135 per month
Total Xpress charges	$275 per month

Frame Relay

It is based on a model whereby either LAN frames or, more commonly, higher-layer packets (such as IP datagrams) are transported through the frame relay network in a "point-to-point" manner using "permanent virtual circuits" (PVCs) to define the two endpoints, eg:

Frame Relay "Committed Information Rates"

Internally, the frame relay network is engineered on the basis that not all nodes will continuously attempt to transmit at their full port speed all of the time. In fact, each port is only guaranteed reliable service at an agreed "Committed Information Rate" (CIR). This is typically less than half of the actual port speed in bps.

The charge for a frame relay PVC is mainly based on the agreed CIR, and to a lesser extent on the actual port speed. To minimise the cost, some providers even allow a CIR of 0 bps.

It is possible for a frame relay user to transmit up to the port speed -- in other words, continuously. However, the network is engineered so that reliable frame delivery becomes less and less probable as the average offered data rate rises above the CIR: ultimately, the network is permitted to drop frames.

Useful links