vishal jain, 200411029 1.1 antnet agent based strategy for cmdr “agent based multiple destination...

38
Vishal Jain , 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Upload: chester-holt

Post on 14-Jan-2016

218 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.1AntNet Agent Based Strategy for CMDR

“Agent Based Multiple Destination

Routing ”

Page 2: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.2AntNet Agent Based Strategy for CMDR

Goals and Non-Goals

Goals To see how simulation helps in studying a system? To see, with an example of AntNet Multicasting, how

modeling is done? To see how by changing system configuration we can study

system in better way? To see how to validate your new idea with the help of

simulation?

Non Goals To study computer network routing in details To see other algorithm for multicasting.

Page 3: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.3AntNet Agent Based Strategy for CMDR

Steps in Sound Simulation

Page 4: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.4AntNet Agent Based Strategy for CMDR

Formulating the Problem:1

Warm up…… What is computer network? Modeling network as graph What is routing between network nodes? What we wish to achieve in routing? A new approach to routing : Multicasting…. Is multicasting is same as multiple point to point routing?

Page 5: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.5AntNet Agent Based Strategy for CMDR

Example

Page 6: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.6AntNet Agent Based Strategy for CMDR

Formulating the Problem:2

Problem statement Let T(s,D) is a tree rooted at source node s with a set of

destination nodes D. Find a tree T(s,D), if it exists, which satisfies the following

optimization problemmin(cost(T))

Delay(PT(s,di)) <= delta for each destination

Bandwidth of each link included in tree must be greater than B.

| Delay(PT(s,di)) - Delay(PT(s,dj)) | <= del for each pair of destination.

Page 7: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.7AntNet Agent Based Strategy for CMDR

How Ants Works To find Shortest Path

Advantages:1. Distributed in nature2. No direct communication is required among agents3. Robust and adaptive to changes

Page 8: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.8AntNet Agent Based Strategy for CMDR

Study and Modelling

Study of “A New Intelligent Agent Based Strategy for Constrained Multiple Destination Routing Problems” , David Elliman and Sherin M. Youssef. The computer Journal, BCS Jan 2004.

Paper Model + True Distributive Approach = New Paper

Page 9: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.9AntNet Agent Based Strategy for CMDR

Brief Conceptual Model of Paper

The forward agent colony searches for the target node it stores path as well as cost of path.

At the destination(s) total tree cost is calculated and then corresponding amount of pheromone is updated at each traversed node in forward journey. This task is done by the backward agent

323.45

NodeCost

Curr Path DelayStart Node

SelectionPool &

Decision

A

B

Page 10: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.10AntNet Agent Based Strategy for CMDR

Is conceptual model Valid?: No

Why? Global Locking Cycle avoiding

New Conceptual Model (Novel Idea © ) Each node will be having information about which colony

has visited it already along with the information from where it arrived to this node. It helps in cycle avoiding as well as agents need not to be LOCKED.

Example:6

3

5

4

0 1 2 3 4 5 6 7 8 9

-1 -1 -1 6 4 6 -1 3 -1 -1

Page 11: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.11AntNet Agent Based Strategy for CMDR

Program Structure:The Network

Page 12: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.12AntNet Agent Based Strategy for CMDR

Program Structure:Node

Page 13: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.13AntNet Agent Based Strategy for CMDR

Make Pilot Run

Page 14: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.14AntNet Agent Based Strategy for CMDR

Tree Cost

Page 15: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.15AntNet Agent Based Strategy for CMDR

Queue at Nodes

Page 16: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.16AntNet Agent Based Strategy for CMDR

Comparison & Conclusion

Page 17: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.17AntNet Agent Based Strategy for CMDR

Page 18: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.18AntNet Agent Based Strategy for CMDR

Successes & Failures

Successes Desired size topology can be constructed at runtime Any value of alpha, beta, gama, delta, del can be given Excellent visualization Results are accurate again

Failure Couldn’t able to separate data and ant generator Traffic move as random not as per the routing table as it

moves in the real network. Note: But to check the feasibility of my idea these

assumptions are ok

Page 19: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.19AntNet Agent Based Strategy for CMDR

Write Computer Program and Verify……

Used OMNet++ as a tool….. Confidence in tool ? Do tool support real life scenario for the simulation ? Degree of insight it gives for the model you are

programming… Support for development (manual , community site, mailing

list etc)….. Developed an application….

Page 20: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.20AntNet Agent Based Strategy for CMDR

Index

Conceptual Framework Black Box Model Understanding Simulation Run Queuing Models and Results

Page 21: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.21AntNet Agent Based Strategy for CMDR

Model

Page 22: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.22AntNet Agent Based Strategy for CMDR

General Node Model

Page 23: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.23AntNet Agent Based Strategy for CMDR

Page 24: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.24AntNet Agent Based Strategy for CMDR

Black Box Model

Queuing ModelStops by statistical data

Arrival Rate

Service Rate

Num Of Server

System Capacity

Server Busy Time

Mean Q Length

Mean Life Time in System

Dropped Customer

Number of Customers

Mean Number of customer present in system

Mean P[ i ]

Page 25: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.25AntNet Agent Based Strategy for CMDR

Features of Implementation

The server is selected for service on the basis of idle time FreeServerList[] Two way send() Methods isEmpty(), recordReq(), Request() Stop by statistical information Generalized Very accurate results

Page 26: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.26AntNet Agent Based Strategy for CMDR

M/M/1

1/ = 0.055555555 1/ = 0.142857142 C=1 System Capacity = inf

Lq = .2474 Lq = .24666

Page 27: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.27AntNet Agent Based Strategy for CMDR

Graph of Mean Q Length

Page 28: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.28AntNet Agent Based Strategy for CMDR

M/M/1/N

1/ = 0.0416666661/ = 0.05C=1System Capacity = 4

Param Book Simulation

Lq 1.494 1.4877

Ls 2.36 2.35329

Wq 0.0863 0.0861177

Ws 0.1363 0.136138

Server Busy Time 0.865596

p3 0.2787 0.250932

Page 29: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.29AntNet Agent Based Strategy for CMDR

Busy Server status

Page 30: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.30AntNet Agent Based Strategy for CMDR

M/M/C

1/ = 0.05555551/ = 0.1666666C=4System Capacity = inf

Param Book Simulation

Lq 1.53 1.52249

Ls 4.53 4.52685

Wq 0.085 0.0847795

Ws 0.252 0.252033

Server Busy Time 3.00436

p0 0.0377 0.038582

What if only three servers are there?

Page 31: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.31AntNet Agent Based Strategy for CMDR

M/M/C

1/ = 0.05555551/ = 0.1666666C=3System Capacity = inf

Param Simulation New Simulation-old

Lq 61.4072 1.52249

Ls 64.3673 4.52685

Wq 3.44632 0.0847795

Ws 3.61261 0.252033

Server Busy Time 2.96004 3.00436

p0 0.00403955 0.038582

Page 32: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.32AntNet Agent Based Strategy for CMDR

Graph

Page 33: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.33AntNet Agent Based Strategy for CMDR

M/M/C/N

1/ = 0.05555551/ = 0.1666666C=3System Capacity = 6

Param Book Simulation

Lq 0.479 0.291294

Ls 3.05 2.99995

Wq 0.271 0.0179682

Ws 1.386 1.66044

Server Busy Time 2.70865

p0 0.0757 0.0496009

Page 34: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.34AntNet Agent Based Strategy for CMDR

M/M/C/C

1/ = 0.05555551/ = 0.1666666C=4System Capacity = 4

Param Applet Simulation

Lq 0 0

Ls 2.3816 2.36732

Wq 0 0

Ws 0.166666 0.166295

Server Busy Time 2.36732

p0 0.061 0.0635515

Page 35: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.35AntNet Agent Based Strategy for CMDR

M/M/inf

1/ = 0.05555551/ = 0.1666666C=infSystem Capacity = inf

Param Applet Simulation

Lq 0 0

Ls 3 2.9935

Wq 0 0

Ws 0.166666 0.166841

Server Busy Time ------- 2.9935

p0 0.0497 0.0506344

Page 36: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.36AntNet Agent Based Strategy for CMDR

M/M/C without & with priority Q

Without Priority With Priority

S[0]=0.72591

S[1]=0.688951

S[3]=0.634994

S[4]=0.572459

S[0] = 0.65387

S[1] = 0.653838

S[3] = 0.653846

S[4] = 0.6539

1/ = 0.05555551/ = 0.1666666C=4System Capacity = inf

Page 37: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.37AntNet Agent Based Strategy for CMDR

Failures

INF=150 or more M/M/inf not satisfactory Dynamic module creation was not successful Calling source limit

Page 38: Vishal Jain, 200411029 1.1 AntNet Agent Based Strategy for CMDR “Agent Based Multiple Destination Routing ”

Vishal Jain , 2004110291.38AntNet Agent Based Strategy for CMDR

What Next with OMNet++

NoC : Async2005 paper for ASD course SCMDR : Antnet Muticast routing for NO course Quantum Cryptography