|
Volume 18, Number 2, 2015
M. PAUN, C. LEANGSUKSUN, R. NASSAR, T. THANAKORNWORAKIJ
Abstract. High Performance Computing applications on Cloud are of significance because of cost-effectiveness and elasticity. Reliability analysis of HPC applications on Cloud is an important area of study to better utilize infrastructure while dealing fault tolerant issues in a Cloud environment. In this work, we present a reliability model of a Cloud system under four scenarios: 1) Hardware components fail independently and software components fail independently; 2) software components fail independently and hardware components are correlated in failure; 3) correlated software failure and independent hardware failure; 4) dependent software and hardware failures. Moreover, we propose an optimal checkpoint placement technique based on reliability information for each scenario. Results show that if failure of the nodes and/or software in the system possesses a degree of dependency, the system becomes less reliable, which means that the failure rate increases and the mean time to failure decreases. Also, an increase in the number of nodes decreases the reliability of the system. Moreover, the optimal checkpoint interval decreases when the reliability of the system decreases. Read the pdf
Voichiţa MAICAN
Abstract. A parallel algorithm for minimum spanning tree and its implementation on a one-chip many-core structure with a MapReduce architecture is presented. The generic structure’s main features and performances are described, but also new general purpose features added for upgrading the existing generic structure in order to perform better in running the proposed algorithms. As the developed algorithm uses the representation of the graph as a matrix, both , dense and sparse cases are considered. A Verilog based simulator is used for evaluation. The main outcome of the presented research is that, compared with the hyper-cube architecture (having p processors and the size in O(p log p)), our MapReduce architecture (with p execution units and the size in O(p)) has the same theoretical time performance: O(NlogN) for p = N = |V|. Also, the actual energy performance of our architecture is 7 pJ for 32-bit integer operation, compared with the ~150 pJ per operation of the current many-cores. Read the pdf
Sakander HAYAT, Mehar Ali MALIK, Muhammad IMRAN Abstract. A topological index can be considered as a transformation of a chemical structure into a real number. The degree based topological indices such as Randić index, geometric-arithmetic (GA) index and atom-bond connectivity (ABC) index are of vital importance among all topological indices. These topological descriptors significantly correlate certain physico-chemical properties of the corresponding chemical compounds. Graph theory has been found to be very useful in this area of research. The topological indices of certain interconnection and mesh derived networks are recently studied by Imran et al. [17]. In this paper, we define some new classes of networks from honeycomb networks by using basic graph operations like stellation, medial and dual of a graph. We derive analytical close formulas of general Randić index Rα(G) (for different values of α) for hexagonal and honeycomb derived networks. We also compute first Zagreb, ABC, and GA indices for these important classes of networks. Read the pdf
Horia-Nicolai TEODORESCU Abstract. Several elementary dynamic properties of cyclic cellular automata are analyzed, with emphasis on the decorrelation of the cycles. Potential applications to the generation of pseudo-random sequences and to spread spectrum communications are briey discussed. Read the pdf
I. ŠIMEČEK, O. MAŘÍK, M. JELÍNEK Le Bail fitting method is a process used in applied crystallography. It can be employed in several phases of crystal structure determination and as it is only one step in a more complex process, it needs to be as fast as possible. This article begins with a short explanation of crystallography terms needed to understand the Le Bail fitting, then continues with the description of the Le Bail fitting method itself and basic principles on which it is based. Then the parallelization method is explained, starting with a more general process, followed by specifics of GPU accelerated computing including short part
Andrei ZENE, Claudiu-Teodor CHIRAP, Octavian CREŢ, Lucia VĂCARIU
Abstract. The NoC domain has known a big development lately, acquiring an ever growing importance in the context of hardware miniaturization. However at this point it is still hard to observe and / or debug what is going on inside the chip, be it for debugging purposes or for long-running processes like computational biology simulations which could gain a great improvement when ran on an FPGA chip. Therefore, this paper proposes hardware implementations of three of the most important snapshot algorithms: Lai-Yang, Li et al. and Mattern, which could be used in order to achieve observability on a long-running process physically implemented on an FPGA target. The setup is based on the layers architecture, making it easy to separate the snapshot algorithm from the application or the intercommunication network. The intercommunication network was generated using the CONNECT NoC generator and the snapshots are sent to a PC via UART for displaying. The algorithms were compared from three points of view: operating frequency, throughput and resource usage. Based on the obtained results, we show that the Mattern algorithm is the best candidate for an effective hardware implementation (both from the resource usage and speed points of view). Read the pdf
Cornel MICLEA, Luminiţa AMARANDE, Marius Cristian CIOANGHER,
Abstract. Water splitting by means of semiconducting photoelectrodes and solar light represents a promising alternative to conventional fossil fuel economy. In this process the photoactive electrode absorb sunlight directly thus initiating the photochemical reaction which create excess electrons in the conduction band of the semiconducting electrode. Titanium doped iron oxide seems to be a promising semiconducting material for photoelectrodes. Consequently, we investigated the effect of Ti doping on the structure, electrical and photoelectrochemical properties α-Fe2O3. The Ti doped α-Fe2O3 were prepared by a slightly modified mixed oxide route, consisting in a prolonged mixing of the raw materials in a high energy planetary ball mill until the particles decreased to the nanometric sizes. Optimum results were obtained for samples doped with 5 at. % titanium and sintered at 1200°C.
|