Impact of networking technology on the access network design

7.2.2.2.1 TDM networks

In TDM networks, which were designed mainly for voice services, an important requirement is to determine the number of voice circuits needed for a given call traffic demand while meeting a certain grade-of-service (GoS). Fortunately, there is an elegant result due to Danish mathematician A. K. Erlang that he developed almost a hundred years ago. In this context, the demand is often referred to as offered traffic or offered load, and is given in the dimensionless unit, Erlang. The traffic offered can be best characterized in Erlangs by the following product:

Offered traffic (a) = Average call arrival rate * average call duration time

Then, for c circuits, the call blocking probability is given by the following Erlang-B loss formula:

E (a,c) = a^c / c!

/

where the summation is from k=0 to c. It may be noted that this formula is developed under the assumption that the arrival traffic follows a Poisson Process, while the result being insensitive to the actual statistical distribution of the call duration time. It is easy to see that this result is applicable to a network link.

Often, we’re interested in determining the number of circuits if the offered traffic and the acceptable grade-of-service (in blocking probability threshold) is given. It is not hard to see that a simple iterative test method can be employed using (7.2.1) so the proper number of circuits required can be determined. A commonly used value for the grade-of-service is 1%

call blocking probability. Thus, for 100 Erl of offered traffic, and for 1% call blocking probability, we can iteratively use Formula (7.2.1) to determine that 117 circuits are needed.

Nowadays, the teletraffic software includes the Erlang-B calculation. The interested reader may also use the freely usable web-based Erlang calculator available at

http://www.erlang.com

In many cases, we’re primarily interested in busy-hour offered traffic and its impact on performance. In many networks, the average Erlang offered traffic per customer can be determined from operational measurements. For example, if average offered traffic per customer is 0.03 Erl, then for an offered traffic of 100 Erl, the average number of customers that can be supported with 117 circuits is a little over 3,300 customers (An astute reader may note that Eng-set model may be more appropriate since a finite population is eventually considered due to 0.03 Erl per subscriber; however, since the population size is large, the Erlang-B loss formula still turns out provide a very good approximation.).

In many practical networks, the offered traffic is hard to estimate. Only, measured traffic, or carried load can be determined. Interestingly, it can be shown that the carried traffic is nothing but the average number of busy circuits. Thus, for a measured carried load a’ and given number of circuits, c, we have the following relation

a’ = a ( 1 – E (a,c) ) (7.2.2)

The above equation can be iteratively solved to determine the offered traffic, a.

To summarize, the advantage of the Erlang-B loss formula is that it gives us the relation between offered traffic, call blocking, and the number of circuits. This model is applicable to any single link design problem; in particular, the formula can be applied for designing access network links, both aggregated wired and wireless.

7.2.2.2.2 ATM Networks

The Asynchronous Transfer Mode (ATM) network is a session oriented cell switching network, capable of carrying many different services. It is based on virtual connections (VC), and each of the virtual connection of the ATM network is characterized by a set of parameters such as maximum required bitrate, burstiness and different quality requirements (QoS) with respect to allowable delay and cell loss rate. In opposite to the IP network, in the ATM

network call (session) admission mechanisms are implemented in order to prevent the

network congestion. A set of services attributes for ATM-based networks has been defined by ITU-T. The following list consist the most important ones in the context of the ATM network design and dimensioning.

• The type traffic, which has to be handled by the network: Constant Bit Rate (CBR), Variable Bit Rate (VBR), etc. When the access network performs the traffic

aggregation of VBR connections the aggregation gain is obtained. If no concentration of the traffic is used or all of the virtual connections are considered as CBR then the TDM dimensioning rules can applied.

• The traffic attributes of ATM networks specify the character of the traffic and can be represented as:

• Peak Cell Rate (PCR)

• Cell Delay Variation Tolerance (CDVT)

• Sustainable Cell Rate (SCR)

• Burst Tolerance (BT)

• Minimum Cell Rate (MCR)

The CBR traffic is described by Peak Cell Rate attribute only.

• Establishment of communication (dynamic, static). This attribute should be taken into account in the access network dimensioning process.

• The traffic symmetry level (traffic: unidirectional, directional symmetric, bi-directional asymmetric).

• Communication configuration (point-to-point, multipoint, broadcast). In most applications only point-to-point links are used.

• Quality of Service (QoS) defines the transport quality of ATM based access networks as well as the availability level. The QoS can be specified by the following

parameters:

• Maximum Cell Transfer Delay (MCTD)

• Mean Cell Transfer Delay (Mean CTD)

• Cell Delay Variation (CDV)

• Cell Loss Rate (CLR)

Initially the ATM technology was used as a core network technology for IP and Frame Relay access/edge networks. At present there are also popular solutions as xDSL, which are native ATM access solutions, thus there is possibility to construct end-to-end native ATM networks.

The methodology, which is used for designing and dimensioning of ATM core networks, can be successfully adopted for broadband, full-service ATM access networks.

7.2.2.2.3 Frame Relay Networks

Frame Relay (FR) is a medium-speed (up to 2 Mbps) connection-oriented packet data transfer service. It uses permanent virtual connections (PVCs) to establish logical connections

between terminals to provide end-to-end FR services. The typical parameters that describe the FR connection are: the interface bit rate, Committed Information Rate (CIR), which describes the guaranteed mean bitrate and the total information transfer delay.

It is a common operators practice to use ATM as a core network for the Frame Relay traffic aggregation and to use ISDN layer 2 technologies in order to reach end-users (HDSL). So, the

design of Frame Relay networks combines the design of the core/edge ATM networks and the classical narrowband design for the ISDN network.

7.2.2.2.4 Native IP networks

The design and dimensioning of IP networks is a troublesome process due to lack of native IP long-range access solutions and the lack of user models. The best effort approach is at present applied to all Internet services. No resource reservation is used and the congestion is a typical behaviour in IP networks. These factors make hard or impossible to use advanced

mathematical tools for IP networks dimensioning. There is however a common opinion that in Internet the access part is the network bottleneck. As Internet access widely used are all existing narrowband networks (PSTN, ISDN) and broadband networks like Frame Relay or ATM (including xDSL) . In opposite to the generic telecommunications architecture in the IP network there is no strict distinction between the access and the core.

Sometimes the placement of edge routers makes such separation.

As a native IP access technology the metropolitan networks (MANs) can be considered. The design and the dimensioning of mesh-like MANs is similar to the design and dimensioning of IP core networks. There are ongoing works on introducing of different class of services using advanced IP network mechanisms (the DiffServ model). At the time of writing of this

document there are no indications related to dimensioning of such a network.