Routing Protocol Evaluation

David Holmerdholmer@jhu.edu

Mobility Models

Random Waypoint Mobility Two parameters

Pause Time (Pt) Max Speed (Vmax)

Each node starts at a random location

Executes loop Pause for Pt seconds Select a random

destination (waypoint) Move to that

destination at a random speed (0,Vmax)

Repeat upon arrival

Random Waypoint Properties Advantages

Easy to implement Allows heterogeneous speeds and temporarily

stationary nodes Disadvantages

Non-uniform node distribution (tend towards center)

Un-stable instantaneous mobility (tends towards zero and oscillates)

Random Waypoint Properties (cont)

Random Waypoint Properties (cont)

Modified Random Waypoint Narrow the random

speed range (.1 Vmax, .9 Vmax)

instead of ( 0, Vmax ) Pre-simulation

mobility Mobility properties

stabilize before routing and data commences

Doesn’t fix non-uniform node distribution

Other Mobility Models Billiard Model

Node selects a random direction, speed, and time Moves in that direction at that speed for that time and then

repeats (may have pause time as well) Bounces off simulation boundary like a “billiard ball” Maintains uniform node distribution, and uniform average

speed (due to time selection) Group mobility patterns

Node mobility is sum of group mobility and individual mobility Used by clustering based routing protocols (well suited for

certain applications like the military) Trace based mobility patterns

Record real life people/vehicle/etc. motion patterns Requires location hardware such as GPS Difficult to try variations or change “parameters”

Routing Performance

Metrics

Routing Protocol Evaluation Metrics Four most common metrics

Delivery Ratio Latency Path Length Optimality Control Overhead

Delivery Ratio Number of packets successfully received by the

destination / number sent by the source Evaluated by setting up a number of “test” flows

in the network Commonly a number of constant bit rate (CBR) flows

with a specified number of packets per second Uses UDP so every dropped packet results in a reduction

of the delivery ratio (no end-to-end retransmissions) Congestion Sensitive

A large enough test load will result in reduced delivery ratio for ANY protocol due to congestion

Mobility Sensitive If the routing protocol does not respond quickly to

topology change, then packets sent on links that no longer exist will be lost

Delivery Ratio Examples

Delivery Ratio vs. Test Load Delivery Ratio vs. Mobility

Latency The time between the creation of a packet and its

delivery to the destination Usually measured using the same setup as

delivery ratio Congestion sensitive

Latency will drastically increase as the congestion limit is reached (due to waiting in large buffers)

Retransmission sensitive Protocols that locally recover packets will achieve higher

delivery ratio but will increase latency On-demand sensitive

Protocols that setup routes after data is sent will have higher latency on the initial packets of a flow

Latency Example

Path Length Optimality The difference between the length of the path

used for sending packets in the protocol and the length of the best possible path

Measurement Protocol path length observed for each packet using test

flows Best possible path computed offline using same mobility

pattern Measure of protocol’s ability to track good routes

Extra hops from non-optimal routes will result in increased congestion and medium utilization

Path Length Optimality Example

Control Overhead Number/size of routing control packets

sent by the protocol Calculated using counters while simulating

with test flows Sometimes expressed as a ratio of control

to data Indication of how efficiently a routing

protocol operates High control overhead may adversely affect

delivery ratio and latency under higher loads

Control Overhead Example