M.A. / A second technological revolution is imminent. As with the first major technological revolution, which was driven by the decentralisation of computing power from that of mainframe networks to that of the “smart” personal-workstation computer, this second revolution is also characterised and driven by decentralised online-communication technology. In terms of this second revolution, however, modern computer users are poised to free themselves of centralised Internet servers for communication and/or file-sharing by connecting to other users directly. This second revolution has even been given a name, that is, “People-to-People” or “Peer-to-Peer computing”, called “P2P computing” for short. Although the basic concept of P2P computing cannot lay claim to being entirely new, it is associated by many a user with file-sharing, particularly with MP3 music files, whilst P2P computing, in fact, constitutes much more, such as instant messaging, group collaboration and distributed computing. Closer to the Information Studies and library environments would be the P2P data replication of a digital-library collection. The extensive use of P2P computing will have a significant impact on the bandwidths available within organisations, to the extent that an unprepared network may very well be overwhelmed by the increased traffic generated by the use of P2P file-sharing. A well-founded understanding of the usage patterns of the individuals using P2P applications will ensure better management and control by the system administrators of and over the expensive information technology (IT) infrastructure. This will also ensure better response to and usage of other systems and applications, such as e-mail and the general World Wide Web utilising the same infrastructure. The principal aim of this study, which was undertaken at the Rand Afrikaans University, is to identify such user behaviour and utilisation patterns in students interacting with P2P systems available on the Internet. The selected respondents, in their capacity of Information Science students, all completed a questionnaire aimed at examining, among other things, aspects such as • information technology (IT) utilisation • the frequency and duration of user sessions • the type of content downloaded and shared • the social aspects associated with P2P file-sharing. P2P computing can be seen as a radically new mindset, setting the stage for a decentralised Internet future. Although distributed computing is dogged by various inherent problems, for example copyright and privacy issues, the most troublesome problem still is that of bandwidth usage. It is to be hoped, therefore, that the application of the research results of this study will ensure the better management of existing infrastructure and resources, as well as the development and rendering of new value-added services within organisations.

With the introduction of the peer-to-peer paradigm in the world of software, a lot of applications have been created in order to such architecture. Most of them are developed for providing a data sharing service among users connected to a network and programs such as Napster, Gnutella, eMule and BitTorrent have became the so called killer-applications. However some eorts have been spent in order to develop other solutions with the usage of peer-to-peer paradigm. In the case of databases some projects are started with the general purpose of sharing data sets with other databases. Generally they push on the idea of providing the data contained in their database schemes with other peers in the network showing concepts such schema matching, mapping tables and others which are necessary to establish connections and data sending. The thesis analyzes some of such projects in order to see which of them is the most dened and well-supported by concepts and deni- tions. Hyperion Project of the University of Torono in collaboration with the University of Trento is the most promising and it aims to be one of the rst Peer-to-Peer Database Management Systems. However the common idea of considering the peer-to-peer paradigm equal to data sharing - in the way presented by applications such as Napster or others - leads to a lot diculties, it is hard to handle the data sets, some operations must be done manually and there can be some cases where the peer-to-peer paradigm is not applied at all. For this reason the goal is to dene and show the concept of peer-to-peer database built from the scratch with a suitable DBMS for it. A real denition of peer-to-peer database has not been ever made and here for the rst time we tried to give one according to our vision. The denition depends on some precise concepts such global schema - which is the original design of the database -, sub-schema - a well logical dened sub-set of entities of the original schema - and binding tables - necessary to allow the creation of constraints and relations among the entities. Then to show the validity of such concepts and how a management system for peer-to-peer databases can be developed and used, a prototype (named Dionysius) has been realized by modifying HSQLDB - an ordinary DBMS developed in Java - and adding the peer-to-peer platform by using the JXTA libray set.

Information Retrieval in P2P environments (P2PIR) has become an active field of research due to the observation that P2P architectures have the potential to become as appealing as traditional centralised architectures. P2P networks are formed with voluntary peers that exchange information and accomplish various tasks. Some of them may be malicious peers spreading untrustworthy resources. However, existing P2PIR systems only focus on finding relevant documents, while trustworthiness of documents and document providers has been ignored. Without prior experience and knowledge about the network, users run the risk to review,download and use untrustworthy documents, even if these documents are relevant. The work presented in this dissertation provide the first integrated framework for trust-aware Information Retrieval in P2P environments, which can retrieve not only relevant but also trustworthy documents. The proposed content trust models extend an existing P2P trust management system, PeerTrust, in the context of P2PIR to compute the trust values of documents and document providers for given queries. A method is proposed to estimate global term statistics which are integrated with existing relevance-based approaches for document ranking and peer selection. Different approaches are explored to find optimal parametersettings in the proposed trust-aware P2PIR systems. Moreover, system architectures and data management protocols are designed to implement the proposed trust-aware P2PIR systems in structured P2P networks. The experimental evaluation demonstrates that P2PIR can benefit from trust-aware P2PIR systems significantly. It can importantly reduce the possibility of untrustworthy documents in the top-ranked result list. The proposed estimated global term statistics can provide acceptable and competitive retrieval accuracy within different P2PIR scenarios.

Includes bibliographical references. / The following calls for research into the ways in which South African university students use technology for peer-to-peer learning (Czerniewicz and Brown, 2005). This study aims to explore the ways in which students mediate one another's learning and the ways in which they use (and don't use) information and communications technology (ICT) to do so. This study provides a snapshot of eleven University of Fort Hare students' peer-to-peer learning strategies. In exploring this phenomenon in context, the social and cultural factors are analysed using Activity Theory, most notably building on the work of Sharples et al (2007). Semi-structured interviews were conducted with students identified as early adopters of ICT for peer-to-peer learning, selected from 375 respondents to the Access and Use Survey questionnaire (Czerniewicz and Brown, 2007). An analytical framework was developed using categories developed from high-level Activity Theory concepts. Sub-categories were developed using key concepts from Sharples, Taylor and Vavoula (2007), Laurillard's (2003) work, as well as, inductively and deductively in relation to the data, following Hardman's (2008) approach. Three key peer-to-peer learning patterns were identified: A group of students preparing for an exam or a test were found to adopt cooperative learning strategies, while students working together towards a tangible output, such as an assignment, adopted collaborative learning strategies. Peer-to-peer tutoring was found to occur where one student is assisted by a moreable peer. Most interviewees' peer-to-peer learning activities were found to occur face-to-face, and the extent and level of interviewees' ICT use was less than anticipated. The findings are not generalisable beyond this small sample, but serve to advance understanding of the processes involved in students' peer-to-peer learning practices.

Peer-to-peer i dagens applikationer har visat sig vara av stor användning. Detta beror på att peer-to-peer, både i teorin och i praktiken, har visat sig vara effektivare i att tillhandahålla data, detta utan att belasta en eller flera servrar med uppgiften. Snabba framsteg i mobil networking, tillsammans med användares begär för fler online tjänster och spel i mobila enheter, är anledningen till varför vi har blivit intresserade av att undersöka mobil P2P. De mål vi hade med projektarbetet var att ta reda på de för- och nackdelar som finns med mobil P2P, samt hur P2P kan implementeras och användas i en applikation. Som ett resultat av arbetet utvecklades en mobilapplikation, som använder en centraliserad P2P struktur för att utföra delar av nätverkskommunikationen. De problem som uppstod med routing och NAT löstes med hjälp av applicering av ICE-protokollet. Den utvecklade applikationen och dess P2P element utvärderades sedan i förhållande till den klassiska server-klient modellen. Systemet visade sig där fördelaktigt, när aspekter som responstid och serverbelastning betraktades. Andra aspekter så som anslutningstid, applikationskomplexitet och stabilitet under normala förhållanden var däremot ogynnsamma. Lösningsstrukturen av den produkt som utvecklades är tänkt att kunna användas av andra utvecklare, som riktlinje eller inspiration för andra produkter. / Peer-to-peer in today’s applications has proved to be of great use. That is because peer-to-peer, both in theory and in practice have proved to be more effective in providing data, without burdening one or several servers with the task. Rapid advances in mobile networking, along with users crave for more online services and games in mobile units, is why we have gotten interested in investigating mobile P2P. The goals of the project were to find out the advantages and disadvantages of mobile P2P, and how P2P could be implemented and used in an application. As a result of the work, a mobile application was developed. The mobile application uses a centralized P2P structure to perform parts of its network communication, and solves problems of routing and NAT through use of the ICE protocol. The developed application and its P2P element was evaluated in relation to the classic server-client model. The system proved advantageous when aspects such as response time and server load was considered. Other aspects, such as connecting time, application complexity and stability, were found inferior under normal conditions. The solution structure of the developed product is thought to possibly be of use for other developers, as guidance or inspiration for other products.

Durant ces dernières années, les délais de transmission sur Internet ont augmenté de manière considérable, causant une détérioration de performances des applications interactives. La cause de ces augmentations de délais est l’excès de mémoire tampon à l’intérieur du réseau, appelé "bufferbloat". Les efforts de recherche dans cette direction vont vers des techniques de gestion des files d’attente actives et des techniques de contrôle de congestion de bout-à-bout. Dans ce contexte, nous avons examiné LEDBAT, un protocole introduit par BitTorrent qui se base sur le délai au niveau transport, et conçu pour transférer grandes quantités de données sans affecter les délais expérimentés par d’autres applications ou utilisateurs. Nous avons analysé la performance de niveau de transport de LEDBAT avec de mesures expérimentales, de simulations et de modèles analytiques, en comparant ses performances au standard TCP ou à d’autre protocoles de failbe priorité. Nous avons ensuite identifié un problème d’iniquité, et nous avons proposé fLEDBAT, qui ré-introduit l’équité intra-protocole. Dans un deuxième temps, nous avons étudié l’impact du protocole LEDBAT sur la performance de BitTorrent. Par des simulations et des expérimentations sur réseaux réelles, nous avons analysé les effets de LEDBAT sur le remplissage des tampons des noeuds d’accès. Les performances de BitTorrent ont été évaluées en termes de temps d’exécution, qui reflète la qualité de l’expérience utilisateur. Dans les deux cas, les résultats ont montré que LEDBAT diminue le temps de traitement par rapport à TCP et réduit de manière significative l’utilisation de tampons, ce qui se traduit par une baisse des délais. / In the last years, Internet delays are considerably growing, causing a performance deterioration of interactive applications. This phenomenon is getting worse with the increasing popularity of bandwidth-intensive applications, as video streaming, remote backup and P2P systems. The cause of these delays has been identified with the excess buffering inside the network, called “bufferbloat”. Research efforts in this direction head toward active queue management techniques and end-to-end congestion control. In this context, we investigated LEDBAT, a low-priority delay-based transport protocol introduced by BitTorrent. This protocol is designed to transfer large amount of data without affecting the delay experienced by other applications or users. First we analysed transport-level performance of LEDBAT using experimental measurement, simulation and analytical model. Specifically, we evaluated LEDBAT as is, comparing its performance to standard TCP or to other low priority protocols. We then identified a later-comer advantage and we proposed fLEDBAT, which re-introduces intra-protocol fairness maintaining the original LEDBAT objectives. Finally we studied the impact of the LEDBAT protocol on BitTorrent performance. Through simulations and real network experiments, we analysed how BitTorrent impacts on the buffer occupancy of the access node. BitTorrent performance was evaluated in terms of completion time, the main metric to assess the user quality of experience. Results showed that LEDBAT decreases the completion time with respect to standard TCP and significantly reduces the buffer occupancy, that translates in lower delays experienced by competing interactive applications.

For traditional web servers, available bandwidth decreases as the number of clients increases. This can cause servers to serve files slowly or to become completely overwhelmed when load grows too high. BitTorrent is a peer-to-peer solution to this problem, but it requires manual configuration for each file to be delivered this way. We develop a new system that integrates peer-to-peer file delivery with traditional client-server downloads. Clients initially attempt to download a file from a web server; if this is too slow, they transition to peer-to-peer delivery. Experiments with a prototype system show that it serves up to 30x faster than traditional web servers.

No description available.

The Peer to Peer Hadoop Distributed File System (P2PHDFS) is designed to store and process extremely large-scale data sets reliably. This is a first attempt implementation of the Statistic Multiplexed Computing Architecture concept proposed by Dr. Shi for the existing Hadoop File System (HDFS) to eliminate all single point failures. Unlike HDFS, in P2PHDFS every node is designed to be equal and behaves as a file system server as well as slave, which enable it to attain higher performance and higher reliability at the same time as the infrastructure up scales. Due to the data intensive nature, a full implementation of P2PHDFS must address CAP Theorem challenges. This MS project is only intended as the ground breaking point using only sequential replication at this time. / Computer and Information Science

Peer-to-Peer (P2P) networks have seen tremendous growth in development and usage in recent times. This attention has brought many developments as well as new challenges to these networks. We will show that agent extensions to P2P networks offer solutions to many problems faced by P2P networks. In this research, an attempt is made to bring together JXTA P2P infrastructure and Jinni, a Prolog based agent engine to form an agent based P2P network. On top of the JXTA, we define simple Java API providing P2P services for agent programming constructs. Jinni is deployed on this JXTA network using an automated code update mechanism. Experiments are conducted on this Jinni/JXTA platform to implement a simple agent communication and data exchange protocol.