the K-Means algorithm for cluster analysis is one of the most influential and popular data mining methods. Its straightforward parallel formulation is well suited for distributed memory systems with reliable interconn...
详细信息
the K-Means algorithm for cluster analysis is one of the most influential and popular data mining methods. Its straightforward parallel formulation is well suited for distributed memory systems with reliable interconnection networks. However, in large-scale geographically distributed systems the straightforward parallel algorithm can be rendered useless by a single communication failure or high latency in communication paths. this work proposes a fully decentralised algorithm (Epidemic K-Means) which does not require global communication and is intrinsically fault tolerant. the proposed distributed K-Means algorithm provides a clustering solution which can approximate the solution of an ideal centralised algorithm over the aggregated data as closely as desired. A comparative performance analysis is carried out against the state of the art distributed K-Means algorithms based on sampling methods. the experimental analysis confirms that the proposed algorithm is a practical and accurate distributed K-Means implementation for networked systems of very large and extreme scale.
the trend toward ubiquitous services and any multimedia, the proliferation of mobile devices and the widespread use of wireless networks imply changes in the design, the implementation and the execution of software ap...
详细信息
the trend toward ubiquitous services and any multimedia, the proliferation of mobile devices and the widespread use of wireless networks imply changes in the design, the implementation and the execution of software applications. Ubiquitous systems are dynamic systems that change their behavior according to user's needs and hardware capabilities at runtime. As it is not desirable to develop these systems from scratch every time, a specific software architecture providing opportunities for dynamic adaptation of systems is necessary. It must be able to create adaptations at runtime in order to provide a dynamic and adaptive behavior for users according to the evolving context. In this paper we present a supervised adaptation platform for applications based components. the CSC platform (Component, Service and Connector) is based on a component/service model that allows adaptation of component-based applications and use service-oriented architecture for providing adaptation services to be embedded in adaptation connectors.
Communication networks that evolve along a Next Generation Network (NGN) concept must provide users withthe ability to communicate and access services regardless of changes in location and technical environment. thes...
详细信息
Communication networks that evolve along a Next Generation Network (NGN) concept must provide users withthe ability to communicate and access services regardless of changes in location and technical environment. these strict requirements impose research challenges on offering Quality of Service (QoS) support for multimedia services. this paper describes simulation and performance evaluation of a mobility-aware session reconfiguration at the application level that maintains and adapts QoS for these services. the approach relies on handling generic mobility events that affect QoS and executing basic operations, called reconfiguration primitives, in response. Proposed functional model is depicted in the scope of ITU-T NGN architecture, along withthe specified signaling procedures. Model simulation and its performance evaluation, which considers scalability issues against number of session participants, are presented in more detail.
We develop and evaluate a rate-allocation multipath algorithm for delivery of video in a Video-on-Demand (VoD) system built upon an overlay network. To evaluate how well the rate-allocation algorithms approach the opt...
详细信息
We develop and evaluate a rate-allocation multipath algorithm for delivery of video in a Video-on-Demand (VoD) system built upon an overlay network. To evaluate how well the rate-allocation algorithms approach the optimum, we have earlier built a benchmarking system that provides the optimal solution for assigning available bandwidth to delivery paths. this solution is also proportionally-fair. We implement a distributed algorithm for multipath rate allocation that operates based on information collected by overlay nodes. the algorithm is evaluated in the OMNet++ simulation environment and analyzed against the aforementioned benchmarking.
this paper presents an online web-based culturally relevant algebra tutor. It is designed for African American students to improve their mathematical skills. As a critical part of a culturally relevant tutor framework...
详细信息
this paper presents an online web-based culturally relevant algebra tutor. It is designed for African American students to improve their mathematical skills. As a critical part of a culturally relevant tutor framework, AADMLSS (African-American distributed Multiple Learning Styles System) focuses on practicing algebra, but can be adapted to other courses. the Algebra Problems Practice (APP) module is the practice environment where students can extract an algebra problem and practice solving it through interaction with a 3D virtual tutor. Practice on this system, tailored with familiar cultural elements, is expected to improve students' ability and skills in mathematics.
Push-to-Talk over Cellular (PoC) is a "walkie-talkie" like voice communication service over mobile wireless networks, based on Voice over IP (VoIP) technology. the traditional PoC defined in the Open Mobile ...
详细信息
Push-to-Talk over Cellular (PoC) is a "walkie-talkie" like voice communication service over mobile wireless networks, based on Voice over IP (VoIP) technology. the traditional PoC defined in the Open Mobile Alliance (OMA) specification is a centralized architecture. In this architecture, servers need to be deployed for the PoC service to work and be widely available, presenting a substantial cost to ISPs. Meanwhile, the problem of network congestion and one point failure existing in the centralized architecture may degrade system performance and cause the failure of an entire system. In addition, the traditional floor control based on the centralized architecture, also suffers from the drawbacks of using a centralized controller. In this article, we propose a full-distributed architecture for PoC application (FDPoC) in data packet voice communication. In the full-distributed architecture, we propose an ALM-based distributed data packet transmission mechanism and a distributed floor control (DFC) algorithm to alleviate the above issues. In terms of average end to end delay, average max end to end delay and average floor determination delay, the performance of the proposed FDPoC and DFC algorithm are investigated through our analytical model.
Peer-to-Peer (P2P) applications running over a distributed hash table (DHT) based overlay do not benefit from dominant characteristics of nodes in the network (such as resources and speed). this paper presents a model...
详细信息
Peer-to-Peer (P2P) applications running over a distributed hash table (DHT) based overlay do not benefit from dominant characteristics of nodes in the network (such as resources and speed). this paper presents a model for applications to set the predominant features of a node and also a way through which it can prioritize those features on several hierarchical layers. Our architecture is based on several hierarchical modules identified by an ID, each one being modeled through a set of rules defined by the application at the upper layer. We provide extensions to the original Chord protocol for scaling under a requested number of interconnected hierarchical modules. the simulation results and theoretical analyses for our model architecture indicate increased performance for scaling and overall routing costs.
distributed Peer to peer systems such as Chord allow peers to perform efficient searches using object identifiers rather than keywords. More specifically, they use a specific structure with some hashing scheme that al...
详细信息
distributed Peer to peer systems such as Chord allow peers to perform efficient searches using object identifiers rather than keywords. More specifically, they use a specific structure with some hashing scheme that allows peers to perform object lookup operations getting in return the address of the node storing the object. Lookups are achieved by following a paththat increasingly progresses to the destination. these systems have been designed to optimize object retrieval by minimizing the number of messages and hops required to retrieve the object. the disadvantage is that they consider only the problem of searching for keys, and thus cannot capture the relevance of the documents stored in the system. this common problem with existing traditional distributed hash table (DHT) is done because they usually ignore the information retrieval algorithms, and thereby rely on keyword based searches. In this paper, we first propose to augment the P2P DHT system Chord with mechanisms for locating data using the information retrieval system LSI to facilitate content-based full-text search in large distributed information systems. Chord-LSI uses latent semantic indexing (LSI) to guide content placement in a Chord such that documents relevant to a query are likely be collocated on a small number of nodes. During a search, Chord-LSI transmit a small amount of data and search a small number of nodes. Simulation results show that Chord-LSI model is 17% more effective than Chord models.
Volunteer computing which benefit from idle cycles of desktop PCs over the Internet can integrate power of hundreds to thousands desktop systems to achieve high computing power. Centralized volunteer computing system ...
详细信息
Volunteer computing which benefit from idle cycles of desktop PCs over the Internet can integrate power of hundreds to thousands desktop systems to achieve high computing power. Centralized volunteer computing system has dedicated servers to maintain information about the resources. However, in the decentralized system resource information is distributed in the system. Resource discovery architecture is a key factor for peer-to-peer based volunteer computing system. Usually, there is a complex relationship between the distribution of resource information and performance of a system. the main contribution of this paper is to develop a proximity-aware resource discovery architecture for peer-to-peer based volunteer computing system. this architecture has simplicity of a centralized system and can achieve close performance compared to this system, and it is scalable. Furthermore, proposed architecture distributes jobs among the resources fairly. Two resource discovery algorithm with and without proximity-aware feature are compared. the proximity-aware resource discovery algorithm can gain a better result.
the study on protein-protein interactions is rapidly increasing; one of the most important findings of such study is the observation of hub proteins that play vital roles in all organisms. Identifying hub proteins may...
详细信息
the study on protein-protein interactions is rapidly increasing; one of the most important findings of such study is the observation of hub proteins that play vital roles in all organisms. Identifying hub proteins may provide more information on essential proteins and lead to more efficient methods for their prediction. Here, we proposed a new network topological-based method for prediction of hub proteins in Saccharomyces cerevisiae (baker's yeast). the method, HP 3 NN (Hub Protein Prediction using Probabilistic Neural Network), has successfully predicts the hub proteins with accuracy of 95% (sensitivity of 1.0 and specificity of 0.89).
暂无评论