# Traffic generation model

(Redirected from Traffic generation)

A traffic generation model is a stochastic model of the traffic flows or data sources in a communication network, for example a cellular network or a computer network. A packet generation model is a traffic generation model of the packet flows or data sources in a packet-switched network. For example, a web traffic model is a model of the data that is sent or received by a user's web-browser. These models are useful during the development of telecommunication technologies, in view to analyse the performance and capacity of various protocols, algorithms and network topologies.

## Application

The network performance can be analysed by network traffic measurement in a testbed network, using a network traffic generator such as iperf, bwping and Mausezahn. The traffic generator sends dummy packets, often with a unique packet identifier, making it possible to keep track of the packet delivery in the network.

Numerical analysis using network simulation is often a less expensive approach.

An analytical approach using queueing theory may be possible for simplified traffic model, but is often too complicated if a realistic traffic model is used.

## The greedy source model

A simplified packet data model is the greedy source model. It may be useful in analyzing the maximum throughput for best-effort traffic (without any quality-of-service guarantees). Many traffic generators are greedy sources.

## Poisson traffic model

Another simplified traditional traffic generation model for circuit-switched data as well as packet data, is the Poisson process, where the number of incoming packets or calls per time unit follows the Poisson distribution. The length of each phone call is typically modelled as an exponential distribution. The number of simultaneously ongoing phone calls follows the Erlang distribution.

## Long-tail traffic models

However, the Poisson traffic model is memoryless, which means that is does not reflect the bursty nature of packet data, also known as the long-range dependency. For a more realistic model, a self-similar process such as the Pareto distribution can be used as a long-tail traffic model.

The actual content of the payload data is typically not modelled, but replaced by dummy packets. However, if the payload data is to be analyzed on the receiver side, for example regarding bit-error rate, a Bernoulli process is often assumed, i.e. a random sequence of independent binary numbers. In this case a channel model reflects channel impairments such as noise, interference and distortion.

## Standardized Internet traffic models

There are at least two standardized traffic generation models for packet-switched wireless networks: the 3GPP2 model and the 802.16 model. The 3GPP2 model is much more complex to implement but it is supposed to give more precise results. The 802.16 model is much simpler in realization.

### 3GPP2 model

The 3GPP2 model is described in.[1] This document describes the following types of traffic generators:

The main idea is to partly implement HTTP, FTP and TCP protocols. For example, an HTTP traffic generator simulates the download of a web-page, consisting of a number of small objects (like images). A TCP stream (that's why TCP generator is a must in this model) is used to download these objects according to HTTP1.0 or HTTP1.1 specifications. These models take into account the details of these protocols' work. The Voice, WAP and Mobile Network Gaming are modelled in a less complicated way.

### 802.16 model

The 802.16 model is much simpler. It was proposed in several 802.16 TG3 contributions.[2] The idea is to define three basic models:

• Interrupted Poisson Process (IPP)
• Interrupted Discreet Process (IDP)
• Interrupted Renewal Process (IRP)
• Virtual Private Network (vps)

and mix them together in order to simulate different kinds of web-traffic. Every interrupted process may be either in ON or OFF state. The packets are generated only in ON state. The lengths of ON and OFF periods, sizes of the packets and intervals between them are defined separately in each model, so these models differ in the way their parameters are defined. These models may be mixed together, for example: 4IPP means a mix of four IPP flows with different parameters. HTTP and FTP is simulated as 4IPP; VoIP is simulated as IDP, 2IDP, 4IDP; Video is simulated as 2IRP.[citation needed]