Data Consistency in the Structured Peer-to-Peer Network
Date Issued
2007
Date
2007
Author(s)
Ou, Cheng-Ying
DOI
en-US
Abstract
There has been a great body of work that proposes mechanisms to enhance the search efficiency of P2P file sharing systems.
Prominent systems such as Chord, Pastry, and Kademlia utilize the concept of distributed hash tables (DHT). These systems, using one of DHT or another, are shown to be efficient at looking up arbitrary files in large-scale P2P networks. To handle churns commonly seen in P2P systems today, most of these systems replicate files and maintain the replicas in accordance to the DHT in use. When the files, or more generally the data, are frequently updated, the data consistency in between replicas becomes a rising problem. To overcome this problem, a number of proposed mechanisms are proposed which aim at keeping the data replicas consistent. To obtain a better understanding of how these kinds of enhancements help improve data consistency, we implement a replica consistency enhancement on a prominent DHT-based peer-to-peer system. Through a set of systematic simulations,
we find that: (1) the availability of data of Chord with the replica consistency enhancement is the same as Chord without the enhancement, even when the churn rate is high; and (2)
Chord without the enhancement and Chord with the replica consistency enhancement have no inconsistent data,
even when the churn rate is high. As the result, the replica consistency enhancement appears to be overkill given its implementation complexity. Our findings in this work provide insights into the making of cost-effective design decisions for P2P mutable file sharing systems.
Subjects
分散式雜湊表
同儕網路檢索系統
同儕網路
資料一致性
peer to peer
P2P
distributed hash table
DHT
data consistency
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-96-R94921046-1.pdf
Size
23.31 KB
Format
Adobe PDF
Checksum
(MD5):f29e023fbd07d8a0f116dcf32a053a49