EEC 289 Q Project Presentation TimeOriented Interlink Locator

Motivation and Limitation ¡ Peer-to-Peer resource sharing among existing network infrastructure is desired ¡

Random Walk ¡ Qin Lv et al. (ACM SIGMETRICS 2002) claims it has the

TOIL Algorithm ¡ Use RTT as the hint for selecting the nodes ¡ Only

TOIL Link Selection A B D D C C F F G E EEC

TOIL Searching Tree A B A D B C C A F E G

Experimental Method ¡ Simulation for 100 nodes and 600 nodes graph ¡ Generated by

Network Topology 100 Node Graph EEC 289 Q 600 Node Graph Wen-Fu Kao, Ying

Performance Comparison Ⅰ 100 Nodes Hit Rate 600 Nodes Hit Rate ¡ TOIL becomes

Performance Comparison Ⅱ 100 Nodes Traffic Load 600 Nodes Traffic Load ¡ Exponential growth

Performance Comparison Ⅲ ¡ Hit Time Ratio (HTR)=T 2/T 1 T 1=E[hit time|walker finds

Conclusion and Future Work ¡ TOIL has the potential; needs further improvement ¡ The

Slides: 12

Download presentation

EEC – 289 Q Project Presentation Time-Oriented Interlink Locator (TOIL) Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang Instructor: Dr. Chen-Nee Chuah

Motivation and Limitation ¡ Peer-to-Peer resource sharing among existing network infrastructure is desired ¡ Request node doesn’t know where the target node is ¡ Caused huge amount of network loading for searching the target EEC 289 Q Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

Random Walk ¡ Qin Lv et al. (ACM SIGMETRICS 2002) claims it has the best performance ¡ Choose successor randomly ¡ Every walker chooses only one link ¡ Stops when TTL=0 EEC 289 Q A B D C F G E Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

TOIL Algorithm ¡ Use RTT as the hint for selecting the nodes ¡ Only search the top three nodes ¡ Distributed, Breadth first ¡ Stops when TTL=0 EEC 289 Q Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

TOIL Link Selection A B D D C C F F G E EEC 289 Q G E Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

TOIL Searching Tree A B A D B C C A F E G A G C F C EEC 289 Q G Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

Experimental Method ¡ Simulation for 100 nodes and 600 nodes graph ¡ Generated by the GT-ITM ¡ Matlab for simulating the k random walker ¡ A C program for simulating TOIL Exhaustive searches all possible path to see the best and the worst case EEC 289 Q Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

Network Topology 100 Node Graph EEC 289 Q 600 Node Graph Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

Performance Comparison Ⅰ 100 Nodes Hit Rate 600 Nodes Hit Rate ¡ TOIL becomes stable when reaching enough TTL EEC 289 Q Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

Performance Comparison Ⅱ 100 Nodes Traffic Load 600 Nodes Traffic Load ¡ Exponential growth vs. Linear growth TOIL is a ternary tree EEC 289 Q Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

Performance Comparison Ⅲ ¡ Hit Time Ratio (HTR)=T 2/T 1 T 1=E[hit time|walker finds the shortest path to the target for a given TTL] T 2=E[hit time|TOIL finds the shortest path to the target for a given TTL] EEC 289 Q 600 Nodes Hit Time Ratio Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang

Conclusion and Future Work ¡ TOIL has the potential; needs further improvement ¡ The performance of TOIL depends on the network topology ¡ Possible modification of TOIL Binary vs. Ternary Tree Replace one of the nodes into a randomly selected one from physically connected neighbors Choose only one node instead of k nodes after certain TTL EEC 289 Q Wen-Fu Kao, Ying Yu Tai, Howard Che-Hao Chang