Proximity Induced Labelling Schemes for Distributed Hash Tables

Show full item record

Title: Proximity Induced Labelling Schemes for Distributed Hash Tables
Author: Warrier, Ajit Chakrapani
Advisors: Dr. Jaewoo Kang, Committee Member
Dr. Khaled Harfoush, Committee Member
Dr. Injong Rhee, Committee Chair
Abstract: P2P systems have been recently introduced as an unconventional approach to networking. Among them, structured P2P systems (or Distributed Hash Tables) have such benefits as load balancing, scalability, and self-organizing nature. Most of the earliest structured P2P systems had virtualized address spaces, hence disregarding underlying physical topologies while creating the overlay. By incorporating knowledge of the underlying topology into the P2P system, efficient overlays can be constructed. There have been several different approaches towards this goal. The most popular approach has been reactive in nature, where nodes having been assigned their virtual identifiers in the overlay, search for good neighbors or routes towards their destination. This work, on the other hand, takes a proactive approach. Our goal is to assign identifiers to nodes so that their position in the overlay would approximately reflect their position in the physical topology. Such identifiers or Proximity Induced Labels would then make the consequent search for good neighbors/routes unnecessary, since they would be implicit by the overlay geometry. We introduce two such labeling techniques, one for the well known Content Addressable Network (CAN), and the other for the binary Hypercube, based on delay information from a set of well-known nodes on the Internet called Landmarks. Our performance evaluation demonstrates that proximity induced labels can be assigned in a scalable manner to CAN without changing the CAN algorithms, leading to better performance than the conventional CAN. Also, such labeling when combined with the high connectivity of the Hypercube, achieves highly efficient overlays at the cost of some increased node state.
Date: 2004-08-16
Degree: MS
Discipline: Computer Science

Files in this item

Files Size Format View
etd.pdf 292.8Kb PDF View/Open

This item appears in the following Collection(s)

Show full item record