Catálogo de publicaciones - libros

This paper proposes an identity authentication mechanism at the link layer for acknowledgment frame in IEEE 802.15.4 network. With the proposed mechanism there are only three bits for authentication, which can greatly reduce overhead. The encrypted bit stream for identity authentication will be transmitted to device by coordinator within association process. Statistical method indicates that our mechanism is successful in handling MAC layer attack.

The illegal distribution of duplicated contents on the Web is causing digital content providers great economic loss. Therefore, Information security is becoming a more important factor in distribution of digital contents. In this study, we designed a digital contents distribution system based on the public key techniques in hierarchical web caching structures. The superior performance of the proposed system has been proven in the experimental tests. The results of experiment show that the supposed system improved the security of DC without decreasing process speed and improved user convenience.

Decentralized, cooperative and large-scale distributed storage systems that consist of a cluster of storage nodes attached with local disks can deliver high resource utilization, high availability and easy scalability. This paper describes the design and prototype implementation of a novel Cluster-Aware Cache (CAC) algorithm that shares memories between nodes in cluster to construct an efficient and cooperative cache-to-disk accesses policy. The difference between our scheme and previous studies is that processes on different node can access the same page concurrently. Furthermore, CAC algorithm is also well suited to heterogeneous clusters where one or more nodes may have larger amounts of memory than the others. The performance measurements with a Web server on our system show dramatic performance improvements with increasing number of nodes.

In an out-of-band SAN virtualization system, the virtualization appliance maintains metadata, and the agents inside the kernel of servers use that data to supply virtual storage devices and to perform the mapping of I/O address. A design of an out-of-band SAN virtualization system based on Windows NT volume manager driver, and its underlining technologies were presented in this paper. It shows that, in general our system is able to supply large volume and high bandwidth virtual storage devices for applications, and it can be used as a basic environment to manage the SAN centrally. The system performance was investigated in comparison with a plain SAN under FAT32 and NTFS, using different data block sizes and access patterns. The results reveal that the overhead induced by our approach is much low. Under FAT32, the performance characteristics of the 3-striped virtual volume follow a typical strip distribution strategy and the bandwidth is 1.20 3.71 times greater than general volume. Furthermore, under NTFS, the bandwidth of the 3-striped virtual volume is an average of 4.10 (max 4.82) times greater than general volume with the random read access test. Hence it can be concluded that our virtualization approach could make use of the storage resources in SAN more effectively.

In this paper, MagicStore, a new out-of-band virtualization system designed for SAN environments is proposed. Online multiplication of the components in a striped volume can help enhance both the I/O performance and storage capacity of a system, but it requires online redistribution of the data on the volume. MagicStore employs a new mapping management solution based on a sliding window to support the online data redistribution without loss of scalability. Furthermore, some virtualization transactions, such as online resizing, require modification of the virtualization metadata, which results in the challenge of keeping the persistent consistency of metadata. MagicStore, by using a combination of ordered writes, REDO logging and log integrity checking, can survive across panics and power failures robustly. In order to support log integrity checking effectively, MagicStore also uses a new log format.

The standard OS and server platform hardware have not been optimized for applications that transfer large multimedia files, resulting in poor server I/O performance. One source of the problem is that several redundant copies are introduces when the data is transferred from disks to a Network Interface Card. To solve the problem of redundant copies, we propose a Contents Delivery Accelerator that accelerates large file transfers by eliminating the redundant copies from disks to the NIC. To eliminate the redundant copies, the CDA introduces a new function, called a logical direct link, which provides the shortest path from the disks to the NIC. By using the shortest path, we can completely eliminate the redundant copies, thereby improving the I/O performance of server. The CDA architecture is a combined hardware-software approach. Thus, it comprises CDA hardware and a modified Linux kernel. We implemented the current version of the CDA on a Linux 2.4.18 kernel and an IXP1200 evaluation board. In the experiment, we compared the logical-direct path with a redundant path. For the transfer of data from disks to the NIC, our experimental results show that the average transfer latency of a direct path is as much as 30 percent less than a redundant path.

This paper addresses the types of QoS dissatisfaction caused by imbalance of the initial I/O workload pattern and storage performance across multiple storage servers in a storage cluster. It next proposes a systematic scheme to resolve the QoS problem that periodically monitors the QoS satisfaction level, analyzes the causes of the QoS problem, and performs data migration based on the analysis result. Finally, it verifies the effectiveness of the proposed scheme under a simulation environment under the different types of QoS dissatisfaction.

A new anonymous secure communication protocol with conditional traceability is proposed to provide personal anonymity and privacy protection, in which a secure mapping function is introduced to provide anonymity and personal information protection, when necessary, only authority principal part can act as arbitrator for communication validation. The proposed protocol has 3 advantages: 1) mutual communication; 2) anonymity of communication. 3) conditional traceability.

In this paper, we introduce a novel architecture for programmable network nodes that work with a large number of real-time video streams. We first discuss challenges in transmission of video streams over bandwidth-limited networks, followed by the active approach as an advance for streaming real-time video. In our model, each programmable node makes admission decision for video frames based on evaluating their potential value. Frames “bid” their expected distortion price and the node chooses the best ones first until resource is fully utilized. Analysis of complexity and overhead shows clear benefit of our framework. Simulation experiments demonstrate its consistent outperformance in comparison to lagrangian-based Rate-Distortion Optimized schemes.

In this paper, we propose a RAID-disk placement algorithm of coded video data and an efficient disk prefetching method to increase the number of clients who can be serviced interactive operations in the media server. Our placement policy is incorporated with a special bit count control method that is based on repeated tuning of quantization parameters to adjust the actual bit count to the target bit count. The encoder using this method can generate coded frames whose sizes are synchronized with the RAID , so that when various fast-forward levels are accessed we can reduce the seek and rotational latency and enhance the disk throughput.