Progress Jan 2011 Putra

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Report Summary

Thesis Submission
Research Summary

Short Update January 2011

Pratama Putra

January 17, 2011

1
Pratama Putra Short Update January 2011
Report Summary
Thesis Submission
Research Summary

Table of contents

1 Report Summary

2 Thesis Submission
Submission Report
Thesis Structure

3 Research Summary
Outline
Layering Algorithm
Experiment Summary

2
Pratama Putra Short Update January 2011
Report Summary
Thesis Submission
Research Summary

Report Summary

Thesis Submission
Submission Report
Thesis Structure
Research Summary
Research Outline
Layering Algorithm
Experiment Result

3
Pratama Putra Short Update January 2011
Report Summary
Submission Report
Thesis Submission
Thesis Structure
Research Summary

Submission Report

Title: “A Distributed Index Poisoning System for Effective


Control of Peer-to-Peer Network Applications”
Submitted on January 14, 2011 23:55 (5 minutes before
deadline)
Volume: 6 Chapters, 109 Pages

4
Pratama Putra Short Update January 2011
Report Summary
Submission Report
Thesis Submission
Thesis Structure
Research Summary

Thesis Structure

Chapter 1: Introduction
Chapter 2: Related Work
Chapter 3: Measuring P2P Network Topology through
Distributed Crawlers
Chapter 4: Characterizing High Influence Using Network
Properties
Chapter 5: Network Structure Based High Influence Peers
Selection for Effective Poisoning
Chapter 6: Conclusion

5
Pratama Putra Short Update January 2011
Research Outline
Background: Problem in P2P Network
High Traffic Poisoning for Controlling Copyrighted File Distribution in P2P

Objective: To Conduct Effective Index Poisoning


Need an Effective Index Poisoning to Reduce the Control Traffic

Approach Solution: Reduce the Index Poisoning Target


Considering Network Structure
Infer Structure from Topology
Layering Algorithm: Define the Most Effective ”Layer” for Index Poisoning

Result
Apply Poisoning to 13% Peers Directly, Ends Up 85% Peers Indirectly
Increase Poisoning Target to 27% Directly, Ends Up 96% Indirectly

Conclusion
Effective Index Poisoning: Reduce Control Traffic by Order of Magnitude
Difference
Enable to Choose the Control Priority: Reduce Traffic or Max Performance
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

The Two Level Layering Algorithm

Level-1 Layering
Divides Network into Layers by Looking at Peers Behavior in
Search Queries:
ORIGIN: Peers that are not Relaying Queries
EDGE: Peers at the Termination of Queries
RELAY: Peers that Relaying Queries between ORIGIN and
EDGE
Level-2 Layering
Divides RELAY by Looking at Peers Proximity with EDGE
and ORIGIN
TOP: Peers Adjacent to EDGE
BOTTOM: Peers Adjacent to ORIGIN
MIDDLE: Others

7
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Bandwidth Distribution in Lv.-1 Layering

Figure: Bandwidth distribution in Lv.-1 Layering

8
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Layering Illustration And Composition

Figure: Layering Composition

Figure: Layering Illustration


9
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Experimental Method

Figure: Poisoning Environment


10
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Evaluation Experiments

Index Poisoning Performance:


1 (Lv.-1) ORIGIN vs. RELAY vs. EDGE
→ Distribute 3 Keys to 1000 peers in each
2 RELAY vs. Random
→ 2 Keys to all RELAY and Random Peers with the same
number
3 (Lv.-2) BOTTOM vs. MIDDLE vs. TOP
→ 3 Keys to all Peers in Each
4 MIDDLE vs. Random vs. Conventional
5 MIDDLE+BOTTOM vs. Conventional

11
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Experiment 1: ORIGIN vs. RELAY vs. EDGE

9
EDGE-Poisoning
RELAY-Poisoning
8

6
Key Spread (Normalized)

0
0 5 10 15 20 25 30 35 40 45
Time (minutes)

Figure: Exp. 1: File keys distributed to the peers in RELAY layer spread
more and faster
12
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Experiment 2: RELAY vs. Random

14000
RELAY(48%)-Poisoning
Random(48%)-Poisoning

12000

10000

8000
Key Spread

6000

4000

2000

0
0 5 10 15 20 25 30 35 40 45
Time (minutes)

Figure: Exp. 2: Similar performance for both methods


13
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Experiment 3: BOTTOM vs. MIDDLE vs. TOP

12000
TOP-Poisoning
MIDDLE-Poisoning
BOTTOM-Poisoning

10000

8000
Key Spread

6000

4000

2000

0
0 5 10 15 20 25 30 35 40 45
Time (minutes)

Figure: Exp. 3: File keys distributed to the peers in MIDDLE layer


spread more and faster, despite of the low ratio.
14
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Experiment 4: MIDDLE vs. Random vs. Conventional

12000
Conventional
MIDDLE
Random

10000

8000
Key Spread

6000

4000

2000

0
0 5 10 15 20 25 30 35 40 45
Time (minutes)

Figure: Exp. 4: File File keys distributed to the peers in MIDDLE layer
spread more and faster compare to random.
15
Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Experiment 5: MIDDLE+BOTTOM vs. Conventional

12000
Conventional
MIDDLE+BOTTOM Poisoning
MIDDLE

10000

8000
Key Spread

6000

4000

2000

0
0 5 10 15 20 25 30 35 40 45
Time (minutes)

Figure: Exp. 5: Almost the same performance with conventional


16
Coverage: up to 90%. Pratama Putra Short Update January 2011
Report Summary Outline
Thesis Submission Layering Algorithm
Research Summary Experiment Summary

Summary

Table: Index Poisoning Method Characteristic


P
Method Performance (P) Traffic Generated (TR) TR
Conventional Very High High +
Level-1 Layering
Very High Medium-High ++
(RELAY)
Level-2 Layering
High Very Low +++
(MIDDLE)
Level-2 Layering
Very High Low +++
(MIDDLE + BOTTOM)

17
Pratama Putra Short Update January 2011

You might also like