Year: 2018
Type: dissertation
Abstract: There is a growing number of electronic smart devices, a decentralized peer-to- peer file storage system can innovate the way we detach from centralized file storage. The objective of developing a completely decentralized peer-to-peer net- work is the significant performance, scalability, cost, and reliability advantages over centralized single site systems. In this paper, we propose a completely decentralized peer-to-peer network model that utilizes state of the art set membership data structures to achieve a bandwidth efficient multi-keyword search. A multi- keyword search allows users to quickly find desired files without directly addressing the files. The network model extends consistent hashing to replicate files quickly across the network allowing for a swift bandwidth efficient multi-keyword search. Our file system is called the Peer-to-Peer Bandwidth Efficient Keyword Search File Storage System (BEKSFSS).This peer-to-peer distributed file system combines a Distributed Hash table with cuckoo filters optimized for high throughput, low latency, and accurate search results. We propose an improvement on an existing P2P file system that utilizesconsistent hashing to replicate files quickly across the network allowing economical multi-keyword searches. Files within the system are replicated across the large P2P network which allows for high availability without bottlenecks of centralized data storage. This infrastructure considers data reliability, availability, and storage overhead through highly distributed file storage.
Author: Justin Lee
Institution: University of California, Santa Cruz
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2940643788
