University of Thessaly (UTH)
Cooperation and fair resource allocation in communication networks
Abstract
dc:descriptionLast few years have witnessed a dramatic growth of communication networks, with peer-to-peer (p2p) overlay communities becoming a new network paradigm for diverse applications and services. Peers can simultaneously act as clients and servers taking part in a collaborative work process. However, although cooperation is of utmost importance to p2p systems, peers seem to be naturally incited only to consume but not contribute to a community. This phenomenon is known as “free riding” and leads to the degradation of the system performance. In this thesis we propose distributed reputation based resource allocation and server selection policies to motivate cooperation among users of p2p networks and provide fairness by guarantying that more contributive peer can receive a better QoS from their community. We use a common system design that can be applied to both single service (e.g., file sharing systems) and multiple services’ systems (e.g., p2p grids). Proposed framework consists of an autonomous reputation scheme, reputation-based resource allocation and server selection policies which are robust under the presence of misbehavior, strategic peers, greedy peers and newcomers. We quantify a peer’s capability in providing a good quality service by a novel reputation metric and an update mechanism is proposed to dynamically track the contribution level of the peer as the system evolves. Our reputation-based allocation policies determine the quality of service given to all competing for resources peers at a given time according to their reputations, demands, and profiles. On the other hand, our reputation-based server selection policy is used to choose suitable providers of the requested service and avoid misbehaving (non contributive) ones. Our simulations in various different setups of peers with different resource capabilities, consumption and service evaluation profiles present how proposed policies outperform previous work both for single and multiple services systems and lead to the dynamic formation of coalitions (cooperation) between peers who mutually benefit by their transactions. We further study systems of peers who use their capacity-limited access links both for their upstream and downstream connections (single capacity limited-link systems). Such peers have to share their available resources between their own and other peers’ needs. In the selfish approach, each peer would like to exploit the full capacity of his access link only for his downloads. However, if all peers acted selfishly, the system would collapse. In order to motivate peers to cooperate under such systems, we combine our distributed reputation-based system with a novel capacity adaptation algorithm. In this way, each peer has to trade off the capacity he will dedicate for uploading in order to increase his reputation and therefore his revenue and the capacity he will dedicate for his downloads. All peers act rationally, trying to maximize their utility. Our proposed policies lead rational peers to cooperation while promoting fairness, as peers receive resources in proportion to their contributions. Our policies outperform existing work in this area in which the slowest link becomes the bottleneck of a heterogeneous system of different link capacity peers. In the contrary, no such bottleneck appears when our policies are used, improving the performance of the system. Moreover, we apply our reputation-based framework in a BitTorrent-like file sharing system and we highlight the performance gains both for ADSL-type and single capacity-limited link systems. Finally, we study a wireless internet sharing community and we propose easily applicable distributed reputation-based policies to motivate users to share their Internet connection in a p2p fashion. The goal is the provision of free and good quality Internet access anytime and anywhere inside the community. We study general systems of users with different contribution and consumption profiles and show that when our reputation-based allocation scheme is applied, mobile users enjoy a QoS Internet connection from their community in proportion to their cooperation level; the more greedy for community resources users are, the more contributive they should be to satisfy their needs. Our allocation scheme is further implemented in 802.11 APs and clients with no modifications at existing standards at the MAC or network layers and experimental results present that the throughputs of the community members are in proportion to their reputations for any channel conditions, type of traffic sent or received, and number of competing clients.
Degree
thesis:*- Grantor dc:publisher
- University of Thessaly (UTH)
- Year dc:date
- 2010
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Satsiou, Anna
- Σάτσιου, Άννα
Subjects
dc:subject × 14- Δίκτυα ομότιμων κόμβων
- Συστήματα βασισμένα στη φήμη
- Συνεργατικά σχήματα
- Δίκαιη κατανομή πόρων
- Δίκτυα επικοινωνιών
- P2P networks
- Reputation-based systems
- Cooperation
- Fair resource allocation
- Communication networks
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/22856