Changes in queries in Gnutella peer-to-peer networks

Abstract

Abstract.Peer-to-peer (P2P) networks have been drawing significant attention in the area of information management and sharing recently. Unlike the World Wide Web (WWW), every peer in a P2P network is both client and server. It hosts information for sharing and submits queries to search for information from all of the other peers at the same time. There are a large number of studies related to the information behavior of WWW search engines in the US and Europe. Several issues have been investigated in these studies including number of unique and repeat queries, length of queries, terms being used in queries, successive searching, multitasking searching, multiple searching sessions, changes of topics, multimedia searching, etc. The number of similar studies in P2P networks is significantly less. We have previously studied the information behavior of Gnutella P2P network based on data collected in 2002. In this paper, we present changes in Gnutella queries, based on the previously collected data and new data collected in 2003. Similar metrics to those used by Spink, Wolfram, Jansen, and Saracevic and in our previous study are used. We found that the number of non-English queries has increased. The number of repeat queries in P2P has decreased but is still more than that for WWW search engines. The length of queries has been increased by about 40%. The topics of interest have shifted from entertainment and sexuality to computers and entertainment. In general, the information behavior of P2P users has been changing as the P2P technology becomes more mature and the P2P users become more familiar with the technology. As the P2P technology begins to be adopted in e-Business, we foresee that there will be more changes in the future.

Implications of the file names and user requested queries on Gnutella performance

“…Analyzing Gnutella queries in December 2000 and January 2001, Sripanidkulchai [26] also observed that 17% of the queries contained non-ASCII characters. Similarly, Yang et al [32] observed that 0.59% of queries issued in July 2002 were not in English. They also observed that by September 2003, 20.85% of the queries were not in English.…”

“…This compares well to the median number of five terms in filenames (Figure 1(a) [33] reported that the average Gnutella object name and query had 8.47 terms and 3.91 terms, respectively. Yang et al [32] reported a median of three and average of 3.74 query terms in July 2002 and a median of four and mean of 5.21 query terms by September 2003.…”

“…Similarly Zhao et al [35] illustrated that Gnutella objects exhibited a Zipf distribution. Yang et al [32] captured Gnutella queries in July 2002 and September 2003. They observed similar behavior to our query data.…”

Web searching in Chinese: A study of a search engine in Hong Kong

“…Analysis on the AltaVista query logs also has been reported (Silverstein, Henzinger, Marais, & Moricz, 1999). Similar studies on peer-to-peer Gnutella network also have been conducted (Kwok & Yang, 2004;Yang & Kwok, 2005); however, most of these studies focused on English search queries only. The information needs and search behaviors of users of non-English search engines can be very different from those of English search engines because of the different nature (e.g., grammar) of these languages and also the different cultures of non-English users.…”

Exploring Users' Information Behavior in Social Networks

“…Yang and Kwok (2005), researched about the changes in the queries made by users on Gnutella (one instance of P2P network) in 2002 and 2003, by saving the data of these queries in a log file and analyzing it using a database management system. They found an increased number of nonEnglish queries, also a decrease on the number of repeat queries but still more than that for WWW search engines, partly because most users know what file they are looking for when they make a query.…”