Time filter

Source Type

Feng Y.,Shenzhen University | Feng Y.,National High Performance Computing Center at Shenzhen | Cao J.,Hong Kong Polytechnic University | Sun Y.,Hong Kong Polytechnic University | And 3 more authors.
Journal of Ambient Intelligence and Humanized Computing | Year: 2010

Service composition is a mechanism to combine two or more services to form a composite service for delivering the users' desirable functionalities. Existing service composition system in pervasive computing does not capture state information of the smart objects (SOs). Based on the study of relationships of SOs' states and services, we identify unqualified composite services generated by existing service composition systems, which are either inefficiently executed or fail to complete their execution. This handicaps the applications of pervasive computing because its applications like health care normally need more reliable and timing services. In this paper, we first formally model SOs' states and their transitions using finite state machines and propose extending existing service description technologies using the SOs'state information. The obtained information is then used in designing an algorithm to compose SOs' services, which avoids generating the identified unqualified composite services. The proof of the efficiency of the composite services obtained by our proposed algorithms is elaborated. Finally, a performance study was conducted to evaluate our algorithm against the one without considering SO state information. Our experimental results show that the composite services generated using our algorithm can execute faster and more reliably. © Springer-Verlag 2010.

Feng Y.,Shenzhen University | Feng Y.,National High Performance Computing Center at Shenzhen | Cao J.,Hong Kong Polytechnic University | Lau I.C.H.,Hong Kong Polytechnic University | And 2 more authors.
International Journal of Parallel, Emergent and Distributed Systems | Year: 2011

Mobile agents have been widely used in distributed computing to take care of the task execution for the user. It is also suitable for mobile and pervasive computing. However, a pervasive computing environment is characterised by high diversity and dynamism, which gives rise to the requirement that a mobile agent executed in such environments has the self-configuring capability. In this paper, we study the problem, when an object providing a service enters the environment, how a mobile agent without prior knowledge about the object can interact with the service. We describe a method for the mobile agent to obtain the corresponding interaction codes and instantiate them for the interaction at runtime with minimal human involvement. We call this component-level self-configuration. Currently, no existing system renders a mobile agent with such capability. We propose a framework, which consists of a unified model for all the participating objects and mechanisms for a mobile agent to dynamically obtain interaction code and self-configuring it for execution. A prototype platform has been implemented and a preliminary performance study has also been carried out. Our experimental results show that the overhead caused by the component-level self-configuration is acceptable. © 2011 Taylor & Francis.

Zhang X.,Henan Polytechnic University | Feng Y.,Shenzhen University | Feng Y.,National High Performance Computing Center at Shenzhen | Feng S.,Shenzhen University | And 2 more authors.
Proceedings - 2011 International Conference on Cloud and Service Computing, CSC 2011 | Year: 2011

Data locality has recently been extensively exploited in Cloud computing to improve system performance. However, when schedule Map tasks in Hadoop MapReduce framework working in a heterogeneous environment, existing methods either cannot reduce the occurrence of these Map tasks or injure fairness, thus degrading the system performance. In order to address this problem, this paper proposes a data locality aware scheduling method to improve the Hadoop MapReduce system performance in heterogeneous computing environments. After receiving a request from a requesting node, our method preferentially schedules the task whose input data is stored on the requesting node. If no such tasks exist, our method will select the task whose input data is nearest to the requesting node, and then make a decision on whether to reserve the task for the node storing the input data or schedule the task to the requesting node by transferring the input data to the requesting node on the fly. As a proof of concept, we implement the method in Hadoop-0.20.2. In order to evaluate the performance, we carry out an experimental comparison study on our proposed method against the default scheduling method used in Hadoop-0.20.2. The experiment results show that our proposed method improves the data locality and reduces the normalized execution time as well as the response time of jobs. © 2011 IEEE.

Hu B.,Shenzhen University | Hu B.,National High Performance Computing Center at Shenzhen | Yang X.,Shenzhen University | Yang X.,National High Performance Computing Center at Shenzhen
Proceedings - 2013 IEEE International Conference on High Performance Computing and Communications, HPCC 2013 and 2013 IEEE International Conference on Embedded and Ubiquitous Computing, EUC 2013 | Year: 2014

The skeletons of the objects in 3D images can be extracted by using 3D image thinning. The application of 3D image thinning for image analysis is hampered by its considerable computation time. By employing the graphics processing unit (GPU), which has tremendous powerful computing power at an incomparable performance-to-cost ratio, the calculation of 3D image thinning can be accelerated. In this paper, we proposed a parallel implementation approach on GPU for the 3D 12-subiteration image thinning algorithm, in which object voxels of 3D image are assigned to threads based on the characteristic of sparse 3D image data. The performance of our approach is analyzed with different image sizes, the ratio of object voxels and the number of thread grids on GPU. The performance of the traditional threads assignment strategy and new threads assignment strategy are compared to show that the proposed approach is more efficient. © 2013 IEEE.

Lu K.Z.,Shenzhen University | Lu K.Z.,National High Performance Computing Center at Shenzhen | Chen G.L.,Shenzhen University | Chen G.L.,National High Performance Computing Center at Shenzhen | And 6 more authors.
Science China Information Sciences | Year: 2010

To eliminate the routing load unbalance among sensor nodes, one approach is to deploy a small number of powerful relay nodes acting as routing nodes in wireless sensor networks, the major optimization objective of which is to minimize the number of relay nodes required. In this paper, we prove that the relay node placement problem in a bounded plane is a P problem, but its computational complexity in general case is quite great. From the geometric cover feature of the relay node placement problem, an O(n 2 log n) time greedy approximation algorithm is proposed, where n is the number of sensor nodes. Particularly, at each stage of this algorithm's iterative process, we first select a critical node from uncovered sensor nodes, and then determine the location of relay node based on the principle of preferring to cover the sensor node closer to the critical node, so as to prevent the emergence of isolated node. Experiment results indicate that our proposed algorithm can generate a near optimum feasible relay node deployment in a very short time, and it outperforms existing algorithms in terms of both the size of relay node deployment and the execution time. © Science China Press and Springer-Verlag Berlin Heidelberg 2010.

Discover hidden collaborations