Hadoop Distributed File System (HDFS) was originally designed to store big files and has been widely used in big-data ecosystem. However, it may suffer from serious performance issues when handling a large number of small files. In this paper, we propose a novel archive system, referred to as Small File Merger (SFM), to solve small file problems in HDFS. The key idea is to combine small files into large ones and build an index for accessing original files. Unlike traditional archive systems such as Hadoop Archives (Har), SFM allows modification of archived files directly without re-archiving. Considering that most of the reads in HDFS are sequential, we design an adaptive readahead strategy based on the Simultaneous Perturbation Stochastic Approximation (SPSA) algorithm to maximize read performance. Furthermore, our system provides an HDFS-compatible interface, which can be used directly without recompiling and redeploying the existing HDFS cluster, hence facilitating convenient deployment for practical use. Preliminary experimental results show that our system achieves better performance than existing methods.

Linked data is a decentralized space of interlinked Resource Description Framework (RDF) graphs that are published, accessed, and manipulated by a multitude of Web agents. Here, we present a multi-agent framework for mining hypothetical semantic relations from linked data, in which the discovery, management, and validation of relations can be carried out independently by different agents. These agents collaborate in relation mining by publishing and exchanging inter-dependent knowledge elements, e.g., hypotheses, evidence, and proofs, giving rise to an evidentiary network that connects and ranks diverse knowledge elements. Simulation results show that the framework is scalable in a multi-agent environment. Real-world applications show that the framework is suitable for interdisciplinary and collaborative relation discovery tasks in social domains.

Previous trust mechanisms are mainly focused on reputational models based on explicit trust ratings. However, the large amount of system-provided and user-generated content published on Web is often ignored. These information is very important for providing more accurate, fine-grained and efficient trust management. In this paper, we propose a content-based trust mechanism and introduce trust reasoning functionalities which can infer a user’s interests and expertise. We present a new method to calculate user similarity based on their feedbacks on reviews to reduce data sparseness of User Similarity matrix. We pinpoint several characteristics of trust which have been overlooked previously: trust is field-dependant, similarity-sensitive, personalized and etc. we differ various trust levels between pairs of users to support more accurate and efficient trust calculation. Although we choose Epinions as our targeted application, the mechanism proposed in this paper are also applicable to other e-commerce systems.

Online travel continues to be one of the most successful e-commerce categories on the Internet. As increasing number of people book online, existing e-ticket systems can not satisfy customers' growing demand. Customers complain about most of the current e-ticket systems for the lack of "reservation" functionality. They also complain that these systems are not "intelligent" enough as a traditional travel agent who can deal with more complicated personal preferences. In this paper, we propose a transaction model for e-commerce system to support reservation functionality while balancing the benefits and risks of both customers and airlines. We introduce the theory of hypothetical query to avoid unnecessary transactions and rollbacks on underlying database. Hypothetical queries can be generated automatically and transformed into relational algebra queries which can be optimized with conventional techniques. We extend the lazy approach for evaluating hypothetical queries with integrity constraints to deal with customers' personal preferences. We also provide the "watch" and "remind" mechanism to help customers to purchase their most desirable tickets. Although we have chosen ticket booking system as our target application, the framework presented in this paper is also applicable to other e-commerce systems.

Trust plays a central role for efficient and secure interactions on the semantic Web. Nowadays, more and more research has focused on this topic. In this paper, we introduce RCSW, a reputation-chain trust model for the semantic Web. We illustrate the basic definitions and mechanisms of RCSW and present the algorithm to calculate trust ratings through the chain of the trust network. RCSW incorporates pairwise trust values and reliable factors of acquaintances and constructs an edge-weighted graph to calculate trust ratings. We also define classes of trust functions to allow users to individualize their trust calculation. Our proposed model can deal with trust management effectively and can quickly adapt to the changing environment.