Earticle

Parallel TCP streams are used for data transfer between clusters in today's high performance applications. When parallel TCP streams are used on LFN, part of streams fail to get enough bandwidth because of congestion generated by the parallel TCP streams themselves [5]. As a result, data transfer throughput is bottlenecked by the slow streams. In this paper, we propose the Stream Harmonizer that schedules packet transmission of parallel TCP streams to balance the bandwidth. The purpose of the Stream Harmonizer is to reduce the congestion that is generated by the parallel TCP streams themselves to improve the performance of the slow streams. The Stream Harmonizer reduces congestion by reducing the bandwidth of fast streams by increasing the round trip time (RTT) of them. A hardware based lightweight scheduling algorithm is used to realize wire-rate intensive scheduling at highspeed backbone network. The Stream Harmonizer is able to schedule packets without exchanging additional information with end hosts. In this paper, we implement the Stream Harmonizer using TGNLE-1, an FPGA (Field Programmable Gate Array) based 10Gbit Ethernet experimental testbed. In addition, we evaluate the Stream Harmonizer using an FPGA based network emulator.

In this paper, we propose a new boosting algorithm for distributed databases. The main idea of the proposed method is to utilize the parallelism of the distributed databases to build an ensemble of classifiers. At each round of the algorithm, each site processes its own data locally, and calculates all needed information. A center site will collect information from all sites and build the global classifier, which is then a classifier in the ensemble. This global classifier is also used by each distributed site to compute required information for the next round. By epeating this process, an ensemble of classifiers, which is almost identical to the one built on the whole data, will be produced from the distributed databases. The experiments were erformed on 5 different datasets from the UCI repository [9]. The experimental results show that the accuracy of the proposed algorithm is almost equal to or higher than the accuracy when pplying boosting algorithm to the whole database.

Recently, multimedia is a more and more important Internet service. For multimedia, Quality of service (QoS) support is a crucial requirement. To meet these QoS requirements, researchers develop specific multimedia mechanism to enhance the performance of video transmission. When they evaluate the performance of their mechanism, most researchers use simulation tools to evaluate. However, when using these simulation tools, researchers usually acquire network-level performance metrics, such as throughput. They cannot evaluate video and audio delivered quality by comparing the original and distorted video. However, using network-level performance metrics can not evaluate the delivered quality correctly. To address this issue, some esearchers proposed simulation tool-sets. These simulation tool-sets can evaluate the video delivered quality well. However, they cannot evaluate the audio delivered quality. Therefore, we propose a new simulation tool-set called as MyEvalvid_RTP to achieve more realistic simulations in this paper. By MyEvalvid_RTP, researchers can evaluate both the video delivered quality and the audio delivered quality.

How to provide low energy consumption and high packet delivery ratio are considered the major issues in the protocol design for the Wireless Sensor Networks (WSNs). The main focus of this paper is to reduce the number of data transmissions within the WSNs for decreasing the energy consumption of the sensor nodes. In the wired networks, the Steiner-Tree is regarded as the optimal approach for constructing the multicast structure for specific senders and receiver. However, the results can not be directly applied to the wireless environment. In this paper, an Energy Conserving Multicast Routing (ECMR) protocol is proposed to reduce the total number of relaying nodes for the construction of a multicast tree. It is designed to be a heuristic scheme since achieving a minimal cost multicast tree is considered an NP-Hard problem in the wireless broadcast environment. Moreover, the proposed algorithm is implemented on practical embedded platforms for performance evaluation. Comparing with the existing multicast routing protocol, the experimental results show that the proposed ECMR scheme can provide better energy conservation while the packet delivery ratio is still preserved.

In this paper, we present a novel feedback control-based load shedding scheme for data stream processing. Firstly we apply system identification to establish a dynamic model to describe data stream management system (DSMS), which enables us to analyze DSMS quantitatively. Then, based on the model, we use the Root Locus method to design the PI controller with proven performance guarantees. Theoretic analysis and experimental results demonstrate that our approach is robust even when system load changes frequently. Comparing to existing strategies, our approach also achieves significantly better performance.

This paper proposes a pitch shifting technique based on the use of Source Filter Model (SFM) by Linear Prediction. A complete solution is described, included estimation of model parameters, pitch detection, Unvoiced/Voiced (U/V) classification and speech synthesis at several pitch values. An automatic U/V classification is proposed using two methods, the Autocorrelation by thresholding and the Zero Crossing Rate (ZCR) versus Energy computation. The satisfactory performance of this technique is evaluated by listening tests using sentences in Standard Arabic Language. The limit range of pitch shifting for a clear speech synthesizing is found.

Pervasive computing environments enable the composition of applications from components allocated across different devices. The applications have to be composed at run- time, to cope with changes in context and resource availability in the environment. In addition, this functionality has to be automated in order to minimize user involvement in application management. We propose two new algorithms which are capable of dynamic allocation of application components to multiple networking devices. These algorithms optimize the selection of the networking evices and the structure of composite applications according to a given criteria, such as minimizing hardware requirements, maximizing the application QoS or other criteria specified by the user. The algorithms are based on generic models. This allows the approach to be used in multiple application domains. We analyze the performance of these algorithms in a simulated nvironment and suggest a system that utilizes our algorithms for pervasive application composition.

In this paper, we propose to support a fixed priority real time scheduler over 802.11 protocol called Deadline Monotonic (DM). We evaluate the performances of this policy for a simple scenario where two stations with different delay constraints contend for the channel. For this scenario a Markov chain based analytical model is proposed. From the mathematical model, we derive the probability distribution of the packet service time at MAC layer. Analytical results are validated by simulation using the ns-2 network simulator.

In case of transmission of fast and protected information, it is important to carry out two operations distinctly. The first, is an operation of encryption, which aims to modify the compressed information to make them unreadable to whoever does not have the necessary authorization to see them (in this paper, we propose give this authorization by using a fingerprint image). The second is an operation of compression which consists in reducing the size of information to be transmitted, by removing redundant information of them. However, these two operations are often carried out separately, although they are strongly bound and one influences on the other. We propose, in this article, to carry out these two operations (compression and encryption) together with a new system able to amalgamate spectral information. This nondestructive spectral fusion, allowing at the same time the compression and the encryption of this information. This system is based on one hand on the rarity of similitude in two different fingerprints and on the second hand on the Discrete Cosine Transformation "DCT", a ransformation used for a long time in JPEG compression.

This paper studies route selection strategy combined with spectrum characteristic in decentralized cognitive radio network. The problem is explored from two aspects: the path stability under spectrum characteristic and interference. Especially, a method that measures the interference to primary users is designed. Under this measurement method, the route selection criteria are explored based on spectrum characteristic. Theoretical analysis and numerical results shows these criteria considering interference is better than random selection.

In this paper, a service architecture based on overlay network was presented. The architecture was designed to build a novel distributed video surveillance system, called as Etouch. In the model, the system is divided into three layers: underlying network layer (UNL) ， overlay service network layer (OSNL) and application layer (AL). OSNL is composed of many intelligent overlay nodes mapped by underlying network layer. Overlay nodes provide all kinds of media service for surveillance application in AL. OSNL allows media service composition of distributed stream processing applications dynamically. And a load balance of QoS-aware service composition algorithm (LBQSC） is employed to satisfy their end-to-end QoS demands. Large-scale experimental results demonstrate the scalability, efficiency and performance of the E-touch system.