CS432 Handouts by VUBuddy

The complex world of networks

Networks are many and diverse. Can we simplify it? Or at least the discussion of it?

Need for NeMS

Let us explore the need and motivation to perform Network Modeling and Simulation (NeMS) by looking at the technology landscape. The landscape consists of the people, technology and their relationships.

Technology Landscape

1. Communications systems: Evolving rapidly
2. User demands: High performance networks
3. Service providers: Rapidly expanding their network infrastructure

Network researchers face the protocol war by developing new communications techniques, architectures and capabilities. Equipment vendors are releasing new devices with increasing capability and complexity. Technology developers and OEMs are developing NG equipment. Network designers and developers are working on how to satisfy the QoS demands of users amidst emerging technologies and techniques viz a viz legacy counterparts? Network Engineer in operations is thinking about what is the right approach to solving problems? Do I buy latest device from company X that claims to solve all my problems? Do I replace underlying technology of my system with the latest generation? Next Generation Network Architect wonders how do I know how this new approach will interact with already existing protocols? How do I build confidence in the utility of this approach without producing and deploying the technology? Is there one solution? Actually not! There are various ways to answer and satisfy the goal seekers. These include

•  Prototyping & empirical testing
•  Trial field deployment
•  Modeling and Simulation (M & S)
•  Analysis

The order represents decreasing costs but increasing abstraction. It is upto the network engineer to trade them off.

What is NeMS?

Network Modeling and Simulation is often considered a single term. In reality, it is not! Simulation is the imitation of behaviour of real-world system or Computational re-enactment according to rules described in model. Whereas modeling is a step that precedes simulations. Together they form an iterative process approximating the real world systems. Model is the logical representation of a complex entity, system, phenomena or a process. In communications, network model could be analytical representation, mathematical form as a state Machine or closed or approximate form. Computer simulation is the execution of computer software that reproduces behavior with a certain degree of accuracy to provide visual insight. It is basically a template on which a computer program runs. It has

•  Inputs
•  Outputs
•  Behaviour

Formally, simulations are pieces of computer software that implement algorithms, take inputs and give outputs.

The model definition could be

•  Descriptive
•  Analytical
•  Mathematical
•  Algorithmic

The computer models can be described into various types

•  Stochastic vs Deterministic
•  Continuous vs discrete
•  Steady state vs Dynamic
•  Local or Distributed
•  Linear or nonlinear
•  Open or closed

These models must be applied according to the perspective. It is important to Model only what you understand. Likewise, understanding your model is equally necessary. Model what you need & no more so that it is neither underdefined nor overdefined.

Simulation Building Process

Consider a one hop communication scenario between two wireless notebooks connected through a WiFi AP. The simulation entities would include wireless computers and their packets (multiple instances), WiFi AP (single instance), and a traffic generator (single instance) that creates wireless computers and their packets. The states of the system would include WiFi AP (idle or busy). Each computer generates a number of packets with each packet successful/failed. The events would include wireless computer creation, packet generation, wireless AP activity. Queues would be needed to schedule the events. These would contain frames waiting in output queue of wireless computer and frames (packets) at WiFi AP input queue. To make the simulation exciting, dynamic and insightful, randomization has to be performed. These would include random realizations of packet lengths, no. of frames per wireless computer, no. of wireless computers, BER (and PER) and packet drop ratio in WiFi AP input queue. It is further needed to distribute various entities and events. The distributions could include Uniform and Gaussian etc for packet lengths and no. of frames per wireless computer.

How Does The Simulation Run?

Following steps are executed during the running of the simulation.

•  Inputs created/initialized
•  Events of transmission, reception and noise occur
•  Randomness causes queues to behave and err
•  Packet successes/failures Simulation logs are compiled and presented as the output in desirable formats

Components of a simulator

•  A self-contained program
•  Event queue
•  Simulation clock
•  State variables
•  Event routines
•  Input routine
•  Report generation routine
•  Initialization routine
•  Main program

Types of simulations

•  Monte Carlo simulation
•  Trace driven
•  Discrete events
•  Continuous events

When to simulate

•  Analytical model not feasible (complex)
•  Analytical model not possible (too simple)
•  Simulate to verify analysis
•  Otherwise simulations are unnecessary