Network simulation
In communication and computer network research, network simulation is a technique where a program models the behavior of a network either by calculating the interaction between the different network entities (hosts/routers, data links, packets, etc) using mathematical formulas, or actually capturing and playing back observations from a production network. The behavior of the network and the various applications and services it supports can then be observed in a test lab; various attributes of the environment can also be modified in a controlled manner to assess how the network would behave under different conditions. When a simulation program is used in conjunction with live applications and services in order to observe end-to-end performance to the user desktop, this technique is also referred to as network emulation.
The simulator (or network simulator) is the program in charge of calculating how the network would behave. Such software may be distributed in source form (software) or packaged in the form of a dedicated hardware appliance. Users can then customize the simulator to fulfill their specific analysis needs. Simulators typically come with support for the most popular protocols in use today, such as IPv4, IPv6, UDP, and TCP.
Open source simulators
The most widely used open source network simulator in research is NS - 2 which runs on Linux. Despite its popularity, a common complaint is the difficulty in learning and using its script based user interface.
Simulations
Most of the commercial simulators are GUI driven, while some network simulators require input scripts or commands (network parameters). The network parameters describe the state of the network (node placement, existing links) and the events (data transmissions, link failures, etc). An important output of simulations are the trace files. Trace files can document every event that occurred in the simulation and are used for analysis. Certain simulators have added functionality of capturing this type of data directly from a functioning production environment, at various times of the day, week, or month, in order to reflect average, worst-case, and best-case conditions. Network simulators can also provide other tools to facilitate visual analysis of trends and potential trouble spots.
Simulation techniques
Most network simulators use discrete event simulation, in which a list of pending "events" is stored, and those events are processed in order, with some events triggering future events -- such as the event of the arrival of a packet at one node triggering the event of the arrival of that packet at a downstream node.
Some network simulation problems, notably those relying on queueing theory, are well suited to Markov chain simulations, in which no list of future events is maintained and the simulation consists of transiting between different system "states" in a memoryless fashion. Markov chain simulation is typically faster but less accurate and flexible than detailed discrete event simulation.Some simulations are cyclic based simulations and these are faster as compared to event based simulations.
Simulation of networks can be a difficult task. For example, if congestion is high, then estimation of the average occupancy is challenging because of high variance. To estimate the likelihood of a buffer overflow in a network, the time required for an accurate answer can be fantastically large. Specialized techniques such as "control variates" and "importance sampling" have been developed to speed simulation.[1][2]