Catálogo de publicaciones - libros

Data integration, in the context of the web, faces new problems, due in particular to the heterogeneity of sources, to the fragmentation of the information and to the absence of a unique way to structure, and view information. In such areas, the traditional paradigms on which database foundations are based (i.e. client/server architecture, few sources containing large information) have to be overcome by new architectures. In this paper we propose a layered P2P architecture for mediator systems. Peers are information nodes which are coordinated by a multi-agent system in order to allow distributed query processing.

Service discovery and location introduces numerous challenges for multi-agent planning in dynamic, real-world domains. Specifically, on non-fault tolerant, peer-to-peer and ad hoc wireless networks, services and agents may become unavailable due to network partitioning, traffic congestion, or attack. Such disruptions might prohibit the ability of agents to find the services needed to execute a plan, possibly threatening the survivability and stability of the overall agent system. This research introduces a method for service discovery and availability prediction based on random walks and demonstrates its applicability in the setting of peer-to-peer, wireless networks.

A Middleware is the software that assists an application to interact or communicate with other applications, networks, hardware, and/or operating systems. We have earlier proposed an RMI-based middleware for mobile devices called System on Mobile Devices (SyD). A middleware on mobile devices is a challenging issue, as it has to deal with problems such as limited memory, frequent disconnections, low bandwidth connection, and limited battery life. The mobile agent module fits in the context of the middleware for mobile devices as it quite naturally alleviates the above mentioned problems. Communication between devices and method invocation capabilities, among other things are carried out by employing agents. In this paper, we provide the design and implementation of an agent module for SyD. We also present practical experiences gathered from carrying out experiments on the agent module.

In this paper we discuss architectural design issues and trade-offs in connection with our experiences porting our agent-based platform, Opal, to the Sharp Zaurus personal digital assistant (PDA). At the present time, the Zaurus is able to run the Java-based Opal platform with RMI, HTTP and JXTA (but not JXME) as message transports. There were many adjustments that had to be made in order to establish JXTA functionality over Java Personal Profile on the Zaurus systems, but it may be an easier process in the future if some of these changes are incorporated into the JXTA standard. The wireless and Bluetooth capability of the Zaurus make it ideal for bridging the gap between Bluetooth networks and traditional networks. The extension of mobility to distributed Agent-based systems will be a significant growth area in future agent research, and the Zaurus PDA a glimpse into the future functionality that mobile distributed agent applications may provide. We also discuss how Opal’s unique support for micro agents may facilitate the deployment of advanced agent systems on future medium- and small-footprint devices.

We propose the ACE (Adaptive Coordinator Election) platform that elects and relocates a coordinator adaptively in P2P networks. In collaborative applications, a coordinator mediates synchronization, consistency, sequencing and delay difference. However, it is difficult to decide a coordinator in applications used in P2P networks because of some characteristics of network, e.g., network instability, and differences in physical networks and devices for participants. The ACE platform elects and relocates a coordinator dynamically according to environmental metrics obtained from participants. Each metric has a priority and weight to allow a coordinator to be chosen according to the purpose of applications. We implemented our platform using the JXTA framework and tested it. The results show the feasibility of adaptive coordinator relocation in P2P networks.

Grid computing aims at integrating geographically distributed computers and providing ‘super-supercomputers’ that can be seamlessly accessed by users all over the world. In peer-to-peer grids, numerous tasks are distributed to grid nodes in a decentralized fashion. In this case, two issues of interest are suitable computing mechanisms and the global performance of the grid, specifically, the dynamics of task distribution and handling. To address these issues, in this paper we present an agent-based adaptive paradigm for peer-to-peer grids and further identify two typical scenarios corresponding to task distribution and handling in this paradigm. We provide two models to characterize the agent-based scenarios. Based on our characterizations, we identify the key features of, and the effects of, several important parameters on the dynamics of task distribution and handling in peer-to-peer grids.

Classical distributed computing projects generally use a specialized client/server model. Recent approaches, such as BOINC, favor instead the development of distributed computing platforms, relying on a generic client/server model. We propose a fully decentralized computing model, considering all participant as peers that can submit personalized computing tasks to any number of other peers currently offering their services, listed in a peer directory. Our model is built upon Chord, a particular Distributed Hash Table. Chord allows load balancing of the number of keys per node, but offers no way to balance the bandwidth load of a frequently accessed key, such as a peer directory. Our model extends Chord with load-balancing of those access-intensive keys. We present a modelization of the bandwidth and storage costs of our model and experimental performance results using a variable number of peers, tasks, tasks time, and a variable ratio of contributors and solicitors roles among peers.

How to Build an efficient Distributed Hash Table (DHT) is a fundamental issue in Peer-to-Peer research field. Previous solutions ignore the heterogeneity of the large scale network. However, in practice, the fact is that the resource held by each peer in the Internet is extremely diverse. And the the willing to share local resources of each peer is also diverse. Therefore, the contribution for the system of a peer should depend on the resources it holds or how many resources it want to share, and should not be uniform. In this paper, we propose a Peer-Performance-Aware Distributed Hash Table (PPADHT) which aims to exploit the heterogeneity. It takes the performance difference of peers into consideration to construct a dynamic variation of wrapped butterfly to achieve the goal. We also show how to optimize the performance of PPADHT in the view of hop counts by random graphs. Our simulation results show that the average lookup hop counts of the PPADHT is approximately a log scale with constant out degrees. And it can achieve loadbalance in two ways: both the document load and message routing load, without introducing any additional load on the peer. Here, the load balance means the load is proportion to the performance of peer.

Systems such as Napster and Gnutella demonstrated the potential of peer-to-peer data sharing. Similar schemes have been used to provide solutions that ensure information availability, survivability and reliability. Current techniques based on Distributed Hash Tables (DHTs) promise scalable solutions for efficient lookup when data is distributed across large networks. This paper considers how to adapt DHTs for use with multi-agent systems, with a goal of supporting distributed data storage and lookup on resource- constrained devices operating on dynamic networks. In existing DHTs, the network and the data are assumed to be static. In our context, sets of mobile agents manage the data. We present a multi-agent approach for building CAN-based DHTs. DHT access is provided through a DHT Agent Service. An extension of the standard CAN lookup algorithm is presented which allows more efficient index maintenance for highly mobile agents. Empirical results verify that the agent-based CAN achieves the expected scalability.

Rapid resource discovery in P2P networks is a challenging problem because users search for different resources at different times, and, nodes and their resources can vary dynamically as nodes join and leave the network. Traditional resource discovery techniques such as flooding generate enormous amounts of traffic, while improved P2P resource discovery mechanisms such as distributed hash tables(DHT) introduce additional overhead for maintaining content hashes on different nodes. In contrast, self-adaptive systems such as ant algorithms provide a suitable paradigm for controlled dissemination of P2P query messages. In this paper, we describe an evolutionary ant algorithm for rapidly discovering resources in a P2P network.