Earticle

Computing environments have been increasingly complex, making managing them difficult. Self-Adaptive methods are used because these do not reveal internal structure and adapt to the environment. Our previous study uses an activation switch that determines components’ tasks. However, they may not work due to the external influences such as viruses because test cases do not contain all cases. If many processes are executing, the shut-down component can waste system resources. In this paper, we propose detecting errors that occur in the Self-Adaptive module and reuse the resources that the shut-down component locks. This can reduce the number of active component when many processes have executed and make the system safer. The paper applies the proposed approach to a video conferencing system to evaluate it. This compares the adaptation of the several approaches with this approach. When many file transfer tasks are active, the number of components is compared. In these experiments, we found abnormal states from the previous study and reduced resource usage through resource reuse.

Networks are becoming much more complicated in their support of ubiquitous computing. For example, broadband mobile communications, wireless ad-hoc networks, IPv6, and so on. Technology is moving towards mobile ad-hoc networks (MANETs), which create temporary networks. The use of mobile devices for composing MANETs is growing, and it is necessary to provide agent service for them. In this paper, we propose a mobile agent platform for supporting MANETs. The proposed algorithms provides agent service among mobile devices, choose an AMS and a DF Service Provider (ADSP) dynamically, maintain the ADSP through another device, and route packets between agents that are unable to communicate. We implement a prototype of these algorithms, based on the Bluetooth protocol. We verify efficient agent management on MANETs, which is not supported by existing mobile agent platforms.

Service-oriented e-science workflow has emerged as a paradigm for integrating heterogeneous distributed science computations. Life sciences also utilizes workflow management systems based on chemical information for accelerating scientific progress. We have developed an infrastructure of chemoinformatics Web services that make those approaches efficient. In this paper, we describe a Web service wrapper for estimating toxic hazard within chemoinformatics workflows. The wrapper makes a secluded application functioned as a Web service process and participated in a series of workflow processes.

Remote sensing images have been fully utilized in disaster detection and other domains. However, many researchers cannot find and access the appropriate remote sensing images as they needed. In this paper, we proposed an effective approach to integrate and share the image resources over the Web. Firstly, the image metadata are exposed based on Grid services and the standard metadata specification; secondly, Agent service is introduced to discover and invoke the metadata services dynamically; thirdly, researchers can query and browse or locate the remote sensing images through the service interface. We have developed a service-oriented remote sensing images integration platform, which supports the parallel query and browse of multi-source remote sensing images. Moreover, it provides better availability and extensibility.

Replication can improve the reliability of a Computational Grid. Replication allows a resource to be used even if some of the in-stances are not running or are inaccessible. However the challenge is how maintain consistent the replicas. In this paper we propose and eval-uate "Oriented Consistency" (OC), a weak consistency algorithm for the dissemination of changes considering application-level in computer Grids. The main aim of this algorithm is optimizing the propagation of changes introducing a preference for nodes in Virtual Organizations of the Grid. The algorithm has been simulated over ns-2, and evaluated with several simple and complex spatial distributions of the users. The impulse response of the algorithm has been characterized. We conclude that considering application parameters such as clustered users of Vir-tual Organizations in the event or change propagation mechanism gives a surprising improvement in the speed of change propagation perceived by most users. In other words, it satis¯es the greatest number of users in the shortest amount of time.

For data gathering in wireless sensor networks, sensors extract useful information from environment; this information has to be routed through several intermediate nodes to reach the destination. How information can effectively disseminate to the destination is one of the most important tasks in sensor networks. Problem arises when intermediate nodes fail to forward incoming packets. Due to limited power and slow processor in each node, algorithms of sensor networks must be designed carefully. Directed Diffusion (DD) is a typical data-centric algorithm which has been used to provide efficient data transmission. We enhance this algorithm based on nodes' credit by using five factors. Simulation results show that our proposed algorithm is more energy efficient and reliable than DD and has the ability of traffic load distribution.

Nowadays, Grid Computing has been accepted as an infrastructure to perform parallel computing in distributed computational resources. Grid has users, resources, and an information service (IS). Resource broker service is one of the main services in grid to find resources, filter resources, allocate resources, etc. Resource selection is part of resource broker that is an important issue in a grid environment where a consumer and a service provider are distributed geographically. In this paper, we design and implement a new data mining –based Grid resource broker service for selection resources on grid environment. The role of this resource broker service is using learning method to find the best nodes according to the requirements of the job and the distributed computing resources on the Grid. The provided application can be executed on top of Globus Toolkit (GT) middleware. The results of experiments show a strong effect in improving resource finding cycle.

Video-on-demand (VOD) is one of the hottest and the most potential services of the broadband network applications. In order to offer the video-information service to numerous users nowadays, in this paper, a load balanced PC-Cluster for the VOD server system has been proposed. We adopt Two-Tier model in the systematic architecture, and use the PC-Cluster to be the storage system of the VOD server. The load balancing mechanism is based on the Least-Connection-First algorithm. Furthermore, a video placement strategy is also proposed in this paper to share and balance the loads among video servers in the cluster. Accompany with the dynamically adjusted files in each video level, a dynamically cyclical video replacement mechanism has been proposed to replicate and allocate video files for improving load balance of the system. The experimental results show that a better load balance of the system can be achieved by applying the proposed scheme.