Search

Rogue peers are a concern because they decrease the utility of the resource-sharing systems, potentially towards to the system collapse. We are currently deploying Peer-to-MANET, a multicast peer-to-peer approach to distribute contents on mobile ad hoc networks. This work presents our efforts to improve trust and fairness, avoiding rogue and selfish peers, in Peer-to-MANET. We employ a modified version of the Network of Favors as a decentralized reputation system to punish rogue peers activities. Through simulations in NS-2, we show that our solution minimizes download success rate by rogue peers and avoid that maliciously peers cause resource depletion of other peers. [ABSTRACT FROM AUTHOR]

Rogue peers are a concern because they decrease the utility of the resource-sharing systems, potentially towards to the system collapse. We are currently deploying Peer-to-MANET, a multicast peer-to-peer approach to distribute contents on mobile ad hoc networks. This work presents our efforts to improve trust and fairness, avoiding rogue and selfish peers, in Peer-to-MANET. We employ a modified version of the Network of Favors as a decentralized reputation system to punish rogue peers activities. Through simulations in NS-2, we show that our solution minimizes download success rate by rogue peers and avoid that maliciously peers cause resource depletion of other peers. [ABSTRACT FROM AUTHOR]

Traditional client/server architectures can deliver high performance if certain configurations are used. Performance quickly deteriorates as the active client base increases unless expensive methods are introduced to scale the system. Research and deployment in the area of peer-to-peer architectures have been engaged in order to address this expense and other problems in traditional architectures (Castro, 2003). BitTorrent is one peer-to-peer architecture that was developed to overcome the mentioned performance deterioration in the file distribution area. It allows peers to participate as servers in the delivery of files while minimizing the need for centralized resources. As no centralized server is used, the initial propagation of the file is determined by how rapidly the file can be replicated in the BitTorrent network. Scalability is determined by how well its algorithms can coordinate and aggregate the resources of peers under heavy loads. A number of measurements were used to assess the performance of the algorithms used in BitTorrent while experiencing a flash mob. The analysis shows that in contrast to traditional client/server architectures, initial flash mobs are not only handled well but encouraged in the BitTorrent environment. [ABSTRACT FROM AUTHOR]

P2P networks were invented with the idea of sharing resources available on the Internet. Their advantages include robustness in failures, extensive resource-sharing, self-organization, load balancing, data persistence, anonymity, etc. Today structured P2P networks are beginning to become more important because they make subsequent queries more efficient and in an internet like scale, query response time is very important. There are two main types of structured overlays: Distributed Hash Table based (DHT) and non-DHT based networks. HyperCup is an example of a non-DHT based structured overlay which is a symmetric overlay based on hypercubes. It has some problems in scalability from the network traffic point of view. DHT-based systems such as Content Addressable Network (CAN) have received a lot of attention and a lot of research has been done on them but still have a few problems, for example the lost indexing problem in CAN. Each of these two types also has its own advantages. In this paper we present a Semantic Addressable Network (SAN) which is a combination of both types. We will focus on CAN and HyperCup. We group nodes in communities based on their main domain ontology concept. SAN helps solve the following problems: semantic unawareness, unnecessary traffic caused by common broadcast methods, and the lost indexing problem in CAN. [ABSTRACT FROM AUTHOR]

Traditional client/server architectures can deliver high performance if certain configurations are used. Performance quickly deteriorates as the active client base increases unless expensive methods are introduced to scale the system. Research and deployment in the area of peer-to-peer architectures have been engaged in order to address this expense and other problems in traditional architectures (Castro, 2003). BitTorrent is one peer-to-peer architecture that was developed to overcome the mentioned performance deterioration in the file distribution area. It allows peers to participate as servers in the delivery of files while minimizing the need for centralized resources. As no centralized server is used, the initial propagation of the file is determined by how rapidly the file can be replicated in the BitTorrent network. Scalability is determined by how well its algorithms can coordinate and aggregate the resources of peers under heavy loads. A number of measurements were used to assess the performance of the algorithms used in BitTorrent while experiencing a flash mob. The analysis shows that in contrast to traditional client/server architectures, initial flash mobs are not only handled well but encouraged in the BitTorrent environment. [ABSTRACT FROM AUTHOR]

P2P networks were invented with the idea of sharing resources available on the Internet. Their advantages include robustness in failures, extensive resource-sharing, self-organization, load balancing, data persistence, anonymity, etc. Today structured P2P networks are beginning to become more important because they make subsequent queries more efficient and in an internet like scale, query response time is very important. There are two main types of structured overlays: Distributed Hash Table based (DHT) and non-DHT based networks. HyperCup is an example of a non-DHT based structured overlay which is a symmetric overlay based on hypercubes. It has some problems in scalability from the network traffic point of view. DHT-based systems such as Content Addressable Network (CAN) have received a lot of attention and a lot of research has been done on them but still have a few problems, for example the lost indexing problem in CAN. Each of these two types also has its own advantages. In this paper we present a Semantic Addressable Network (SAN) which is a combination of both types. We will focus on CAN and HyperCup. We group nodes in communities based on their main domain ontology concept. SAN helps solve the following problems: semantic unawareness, unnecessary traffic caused by common broadcast methods, and the lost indexing problem in CAN. [ABSTRACT FROM AUTHOR]