Catálogo de publicaciones - libros

DHT(Distributed Hash Table) provides a very effective and reliable search scheme in P2P networks. However, when the search involves a multiple-keyword query, it suffers heavy network traffic due to the passing around of a large inverted list among P2P nodes. In this paper, we propose Distance-Based Pruning technique to cut down the size of the inverted list considerably. It utilizes the concept of distance between keywords in the query and removes those entries in the inverted list that are going to be dropped sooner or later. We prove this prediction is accurate and effective such that it reduces the size of the inverted list by 29%.

With the enrichment of digital content over the Internet, more and more people are using software for content sharing especially for multimedia content sharing, and most of these softwares are on the basis of P2P architecture which makes good use of high scalability and high efficiency. However, most P2P file sharing systems do not have security mechanism to protect intellectual property of authors and issuers. Moreover, it is difficult to control each peer’s behavior due to its self-control principle and light-weight server. So we propose a new system based on P2P architecture for file sharing, to which we apply new DRM(Digital Right Management) mechanisms. Results have shown that our system outperforms other systems in features of security, server load and scalability.

Current representatives of Grid systems are Globus and Web Services, however, they have poor scalability and single point failure. It is a hot research topic to build a P2P and grid hybrid framework for resource management and task schedule. We propose Information Pool Based Grid Architecture (IPBGA), which is a real sense hybrid of P2P and grid instead of only introducing P2P methods into grid systems for resource management. The key of IPBGA, information pool protocol, is presented. In our information pool protocol, all of resources and tasks are abstracted into information, and resource requests for tasks and task appeals for resources are viewed as information services, then grid resource management and task schedule are treated as information matching. Therefore, our architecture is very adaptive to heterogeneous, dynamic, and distributed grid systems. We use tri-information centers for collecting information, which strengthens the robustness of our system. Simulation experiments show information pool protocol of IPBGA is more efficient in resource management and task schedule, and has less bandwidth and processing cost compared with other hybrid P2P systems.

In this paper, we present a game-based model to analyze nodes’ behaviors and influence of incentive mechanism on nodes in a peer-to-peer network in which the altruistic and selfish peers coexist. In this model, a mental cost is attached to a peer to describe the level of the peer’s altruism and determine the type of the peer. The merit of our model is the relation between the equilibrium and incentive mechanism(i.e. the influence of incentive mechanism on equilibrium) can be denoted by an analytic formula directly. Therefore, we can evaluate and compare the efficiency of diversified incentive mechanisms conveniently, the network operators can choose proper incentive mechanism to achieve certain optimal objective.

We propose a source peer selection algorithm which efficiently chooses peers as sources from a set of peers having the requested file in a hybrid P2P file sharing system. Our proposed algorithm considers the energy factors as well as the communication issues and the peers’ contribution to the file sharing system. The core idea of our algorithm is to minimize the workload of the wireless peers, and to leave those jobs to cable connected peers which are much stronger than wireless ones. We validate our algorithm through comprehensive simulations, which show that the algorithm works correctly and effectively.

While simulating a P2P system with distributed simulator, it generally requires that one single large network topology should be pre-divided into some small sub-nets, each of which denotes a group of peers in the P2P system. Because of interconnectivity of the simulated network, the sub-nets running on different simulation node must exchange message with each other to complete the simulation task. Based on the knowledge of degree sequence and breadth-first search, this paper proposes a novel approximate algorithm of k-graph partition. By this optimized algorithm, a large P2P network topology can be divided into k sub-nets while not only the traffic among different sub-net is minimized, but also the tasks of simulation are balanced. Through the analysis of time complexity, load balance and edge-cut experimental results for different network topology, it shows our algorithm is a feasible method applied for distributed P2P simulation systems.

Incremental computing masks the communication latency by overlapping computations with communications. However, a sequence of messages with a large latency variance still makes computations proceed intermittently. It is known that a input stream from a data server maximizes the CPU utilization of the networked computation server [7]. Unfortunately, the problem of finding a dominant input stream is -hard in the strong sense. In this paper, a dominant input stream for LU decomposition is proposed. It is shown that the dominant input stream outperforms the input stream sending data in traditional order. In addition, the nonexistence of dominant input streams is proved for the case that the compressed format is used for sending input data.

RAID, as a popular technology to improve the performance and reliability of storage system, has been used widely in computer industry. Recently, the technique of designing data layout in order to fit the requirements of networked storage is becoming a new challenge in this field. In this paper, we present a double-objective Genetic Algorithm for parity declustering optimization in networked RAID with a modified NSGA, we also take and as two objects to find optimal data layout for parity declustering in networked RAID.

Several diffusion schemes have been developed for load balancing on general networks. But these schemes are not well adapt to the optical transpose interconnection network (OTIS) because this network usually has high order Laplace matrix such that computing its spectrum becomes complicated. Even if its spectrum is obtained simply, diffusion schemes sometimes are rather difficult to implement because of large scale of network usually.

Corresponding to traditional X schemes, we propose hybrid diffusion schemes called DED-X for load balancing on OTIS network. By DED-X schemes, load flows are scheduled on intragroup and intergroup links on OTIS network separately. The DED-X schemes only compute the Laplace spectrum of the factor graph of the OTIS network. The spectral information of whole OTIS network is not necessary. We also provide some theoretical evidences to show that DED-X schemes are better than those traditional X schemes. Simulation results show that proposed schemes have significant promotion in efficiency and stability.

The problem of efficiently disseminating data from a mobile stimulus (source) to multi-ple mobile users (sinks) in a large-scale sensor network has been challenging. We address this prob-lem by proposing a minimum-energy tree-based data dissemination scheme, (DLATS). We exploit the fact that sensor nodes are stationary and location-aware. In DLATS, each sensor node finds its (RNG) neighbors and a RNG is generated over the whole network. Then, the source broadcasts its position information to all the other nodes using our localized minimum-energy broadcast protocol, (IRBOP). A dynamic agent tree is generated between each source and multiple sinks using our (SDPF) where the sinks become the agents, e.g., the leaves of the agent tree. Finally, each source uses IRBOP for multicast-ing the stimulus data to the users over the agent tree. We evaluate the performance of DLATS through simulations. Results show DLATS outperforms previously proposed protocols for data dis-semination in large-scale sensor networks.