You are on page 1of 22

ENSC 833-3: NETWORK PROTOCOLS AND PERFORMANCE

CMPT 885-3: SPECIAL TOPICS: HIGH-PERFORMANCE NETWORKS

FINAL PROJECT PRESENTATIONS

Fuzzy Logic Routing in OPNET 9.0

Edward K. Chen
97300-2328
ekchen@cs.sfu.ca
ROAD MAP
u Introduction

u Project Overview
• Fuzzy Logic
u What is it?
u Where is it used?

• Project Scenario and Setup

u Simulation Results and Comparison

u References and Questions


INTRODUCTION

u Explosive growth and use of Internet in past decade

u Saturation of network resources

u Need to effectively route packets to minimize end-to-end


delay

u Utilization of resources à minimize buffer size with same


arrival/service rates

u Fuzzy Logic based router in simple application


Project Overview
u Node Model
Project Overview
u Objective:
• Minimize end-to-end delay of packets
• Minimize buffer size

u Decision made at hub to send packets to one of two servers

u Ability to make “smart” decision to benefit overall system,


not individual packets.

u Example:
• Packet “A” joins Queue_1 with less packets
• If ? 1 >> ? 0 : longer overall system delay
• When no additional packets à individual = system
performance
Project Overview

u Application of Fuzzy Logic to achieve objectives

u Other Routing Algorithms considered


• Shortest Queue
u Packets sent to queues with least number of customer

• Round Robin
u Packets sent to alternate queues
Fuzzy Logic

u Introduced by Prof Lotfi Zadeh of UC Berkeley in the 1960s

u Closely resembles human reasoning, uncertainty and vagueness

u Used in control application: refrigerators, washing machines,


welding machines, cameras and robots

u Classic Logic has two value: true or false

u Real life systems are non-linear: classic logic not adequate

u Example:
• How are you feeling today?
• How was the movie?
Fuzzy Logic Component Description

u Four main components:

• Fuzzification:
u Input converted to set of certainties according to
membership functions

• Fuzzy Rule-Base:
u Set of IF-THEN rules used to decide outcome
u Designer dependent à different conclusions can be drawn
from same data
Fuzzy Logic Component Description

• Inference Mechanism:
u Two Step Process
• Determine which rules from the IF-THEN rules apply
• Determine which conclusion reached based on rules

• De-Fuzzification:
u Converts output of Inference Mechanism into a quantitative
number
Fuzzy Logic Component Diagram

Fuzzy Fuzzy
Inputs Inference Output
Fuzzification De-Fuzzification
Mechanism

Rule_Base
Project Description
u Hub Process Model
Project Description

• Two States needed to properly route packets

• Challenge: Implementation of Fuzzy Logic


within the hub

• IDLE
u Initialization of variables.
u No init state used coz all variables needs to be reset
each time

u Route PKT
• Uses fuzzy logic to route packets
• Dependent on parameters: serve rice rates (2), queue
lengths(2) and arrival rates (3)
Project Description

u Fuzzification:

• Two Inputs:
#1 Expected-delay-difference

 1   1 
s = (s 0 + 1 )  − (s 1 + 1 ) 
 µ 0   µ1 

 1 
where (s 0 + 1 )  is expected delay of Queue_0
 µ 0 

#2 Arrival Rates of three input streams:


• ? 0, ? 1, ? 2
Project Description
u Membership Functions
• Expected_Delay_Difference

• Arrival Rates
Project Description
• Rules “Active” when mui > 0
u If Expected_Delay_Diff = 0.5, ? rates all equal 3.5

u NS, ZO, PS “active” for Expected_Delay_Diff


u PS, PB “active” for arrival rates

u 3 x 2 x 2 x 2 = 24 rules active
Project Description
u Inference Mechanism:

• Determine which conclusion reached based on


rule_base.

• Total: 5 x 3 x 3 x 3 = 135 rules

• Rules are designer-dependent

• Different rule yield different result

• Combination of results sent to Defuzzification to decode


u 24 results in Example
Project Description

u DeFuzzification:

• Converts multiple results into single quantitative figure

• Uses Center of Gravity (COG) Method

output =
∑ b ∫ µ (i )
i i

∑ ∫ µ (i )
i

where bi = max. value of membership function


µ (i ) = area of the membership function with
∫ specified certainty
Results and Simulation

u Parameters:
? 0=0.5 packet/sec µ 0=2 packet/sec
? 1=0.3 packet/sec µ 1=1 packet/sec
? 2=0.6 packet/sec
Results and Simulation
Results and Simulation
Future Work

u Adaptation of Fuzzy Logic Router to more complex system.

u Use Fuzzy Logic Control as a general approach to more practical


situations

u Networks with finite buffers and non-Markovian parameters

u Ability to prioritize different packets (guaranteed QoS)


References
u [1] J. Walrand and P. Varaiya, High-performance Communication Networks,
Second edition, Morgan Kaufmann, 2000

u [2] R.Cheng,
R.Cheng, C.Chang,
C.Chang, “Design of a fuzzy traffic controller for ATM networks” IEEE/ACM
IEEE/ACM
Transactions, Volume: 4 Issue: 3 , Jun 1996 Page(s): 460 - 469

u [3] M.Salamah,
M.Salamah, H.Lababidi,
H.Lababidi, “FBLLB: a fuzzy-
fuzzy-based traffic policing mechanism for ATM
networks”, ACS/IEEE International Conference 2001,Volume: 3, Issue:2,
Issue:2, Page(s):
Page(s): 31 -
35
[4] A.Kasiolas,
A.Kasiolas, D.Makrakis,
D.Makrakis, “A fuzzy-
fuzzy-based traffic controller for high-
high-speed ATM networks
using realistic traffic models “ Multimedia Computing and Systems,.
Systems,. IEEE Transactions ,
Volume: 2 , Jul 1999 Page(s): 389 -394

u [5] R.Zhang,
R.Zhang, Y.Phillis,
Y.Phillis, “Admission control and scheduling in simple series parallel networks
networks
using fuzzy logic”, Fuzzy Systems, IEEE Transactions, Volume: 9 Issue: 2 , Apr 2001
Page(s):
Page(s): 307 – 314

u [6] L.Barolli,
L.Barolli, A.Koyama,
A.Koyama, T.Yamada,
T.Yamada, S.Yokoyama,
S.Yokoyama, T.Suganuma,
T.Suganuma, N.Shiratori,
N.Shiratori, “A fuzzy
admission control scheme for high-
high-speed networks”, Proceedings of 12th International
Workshop on Database and Expert Systems Application, 2001. Pages 157-
157-161.

u [7] “What is Fuzzy Logic” Pacific Northwest National Laboratory


http://www.emsl.pnl.gov:2080/proj/neuron/fuzzy/what.html (March 26, 2003)

u [8] D.Mitchell, J.Yeung “Implementation of Start-time Fair Queuing Algorithm in OPNET”


Ensc835, School of Engineering Science, Simon Fraser University, April 2002

You might also like