Showing papers on "Spark (mathematics) published in 2013"

Memory or Time: Performance Evaluation for Iterative Operation on Hadoop and Spark

Abstract: Hadoop is a very popular general purpose framework for many different classes of data-intensive applications. However, it is not good for iterative operations because of the cost paid for the data reloading from disk at each iteration. As an emerging framework, Spark, which is designed to have a global cache mechanism, can achieve better performance in response time since the in-memory access over the distributed machines of cluster will proceed during the entire iterative process. Although the performance on time has been evaluated for Spark over Hadoop, the memory consumption, another system performance criteria, is not deeply analyzed in the literature. In this work, we conducted extensive experiments for iterative operations to compare the performance in both time and memory cost between Hadoop and Spark. We found that although Spark is in general faster than Hadoop in iterative operations, it has to pay for more memory consumption. Also, its speed advantage is weakened at the moment when the memory is not sufficient enough to store newly created intermediate results.

Log analysis in cloud computing environment with Hadoop and Spark

Abstract: Log is the main source of the system operation status, user behavior analysis etc. Log analysis system needs not only the massive and stable data processing ability but also the adaptation to a variety of scenarios under the requirement of efficiency, which can't be achieved from standalone analysis tools or even single cloud computing framework. We present a unified cloud platform for batch log data analysis with the combination of Hadoop and Spark. Hadoop provides a distributed file system and off-line batch computing framework, while the computing pattern in Spark is based on distributed memory. The joint of Hadoop, Spark and the data warehouse and analysis tools of Hive and Shark makes it possible to provide a unified cloud platform with batch analysis and in-memory computing capacity in order to process log in a high available, stable and efficient way.

Quasi-Dimensional Simulation of Spark Ignition Engines: From Thermodynamic Optimization to Cyclic Variability

Abstract: Based on the simulations developed in research groups over the past years, Introduction to Quasi-dimensional Simulation of Spark Ignition Engines provides a compilation of the main ingredients necessary to build up a quasi-dimensional computer simulation scheme. Quasi-dimensional computer simulation of spark ignition engines is a powerful but affordable tool which obtains realistic estimations of a wide variety of variables for a simulated engine keeping insight the basic physical and chemical processes involved in the real evolution of an automotive engine. With low computational costs, it can optimize the design and operation of spark ignition engines as well as it allows to analyze cycle-to-cycle fluctuations. Including details about the structure of a complete simulation scheme, information about what kind of information can be obtained, and comparisons of the simulation results with experiments, Introduction to Quasi-dimensional Simulation of Spark Ignition Engines offers a thorough guide of this technique. Advanced undergraduates and postgraduates as well as researchers in government and industry in all areas related to applied physics and mechanical and automotive engineering can apply these tools to simulate cyclic variability, potentially leading to new design and control alternatives for lowering emissions and expanding the actual operation limits of spark ignition engines

Sparkler: supporting large-scale matrix factorization

Abstract: Low-rank matrix factorization has recently been applied with great success on matrix completion problems for applications like recommendation systems, link predictions for social networks, and click prediction for web search. However, as this approach is applied to increasingly larger datasets, such as those encountered in web-scale recommender systems like Netflix and Pandora, the data management aspects quickly become challenging and form a road-block. In this paper, we introduce a system called Sparkler to solve such large instances of low rank matrix factorizations. Sparkler extends Spark, an existing platform for running parallel iterative algorithms on datasets that fit in the aggregate main memory of a cluster. Sparkler supports distributed stochastic gradient descent as an approach to solving the factorization problem -- an iterative technique that has been shown to perform very well in practice. We identify the shortfalls of Spark in solving large matrix factorization problems, especially when running on the cloud, and solve this by introducing a novel abstraction called "Carousel Maps" (CMs). CMs are well suited to storing large matrices in the aggregate memory of a cluster and can efficiently support the operations performed on them during distributed stochastic gradient descent. We describe the design, implementation, and the use of CMs in Sparkler programs. Through a variety of experiments, we demonstrate that Sparkler is faster than Spark by 4x to 21x, with bigger advantages for larger problems. Equally importantly, we show that this can be done without imposing any changes to the ease of programming. We argue that Sparkler provides a convenient and efficient extension to Spark for solving matrix factorization problems on very large datasets.

Controlled Spark Ignited Flame Kernel Flow

Abstract: In some aspects, a spark plug includes a spark gap in an enclosure of the spark plug. The spark plug includes a passage in the interior of the enclosure. During operation of the engine, the passage directs flow through the spark gap, primarily away from a combustion chamber end of the enclosure. The passage can direct flow at a velocity of 5 meters/second or greater.

Homogeneous charge compression ignition technology implemented in a hybrid electric vehicle: System optimal design and benefit analysis for a power-split architecture

Abstract: Homogeneous charge compression ignition technology can improve fuel economy by providing increased efficiency at low-load operation. This article examines the implementation of this technology in hybrid propulsion systems. To assess the benefits, a physics-based model for a spark ignition–homogeneous charge compression ignition dual-operation engine is developed, together with system and component models, and is used to optimize a crossover sport utility van with a power-split hybrid powertrain. Comparison of optimal designs for the pure spark ignition and dual homogeneous charge compression ignition cases indicates the reduction in the fuel consumption based on our modeling assumptions to be in the range 2.5–5%, depending on the test cycle. These benefits increase substantially when the acceleration performance requirements increase. An analysis method is presented to show how such engine-level changes affect the entire powertrain characteristics, and mode maps are developed to indicate when the benefits...

Sparse Auditory Reproducing Kernel (SPARK) Features for Noise-Robust Speech and Speaker Recognition

Abstract: The speech feature extraction algorithm is based on a hierarchical combination of auditory similarity and pooling functions. Computationally efficient features referred to as “Sparse Auditory Reproducing Kernel” (SPARK) coefficients are extracted under the hypothesis that the noise-robust information in speech signal is embedded in a reproducing kernel Hilbert space (RKHS) spanned by overcomplete, nonlinear, and time-shifted gammatone basis functions. The feature extraction algorithm first involves computing kernel based similarity between the speech signal and the time-shifted gammatone functions, followed by feature pruning using a simple pooling technique (“MAX” operation). Different hyper-parameters and kernel functions may be used to enhance the performance of a SPARK based speech recognizer.

Transition of a diffuse discharge to a spark at nanosecond breakdown of high-pressure nitrogen and air in a nonuniform electric field

Abstract: The transition of a runaway-electron-induced diffuse discharge initiated in a nonuniform electric field under a high pressure of air and nitrogen to a spark is studied. High-voltage pulses with a rise time of 0.5 ns are applied to a discharge gap with a tubular cathode having a small radius of curvature. It is shown that the leader of the spark discharge propagates toward the tubular cathode along preproduced tracks and may pass from one track to another. For a pulse rise time of about 0.5 ns and a gap length of 12 mm or more, it is found that spark leaders originating at the cathode (which has a small radius of curvature) do not reach the anode and accordingly, do not cause the spark breakdown of the gap. It is confirmed that the spark breakdown of the gap is associated with a spark leader that moves away from the plane electrode after the appearance of a bright spot on it.

A focused electric spark source for non-contact stress wave excitation in solids

Abstract: A focused electric spark is used as a non-contact acoustic source to excite stress waves in solids. The source consists of an electric spark source located at the near focus of an ellipsoidal reflector that focuses the acoustic disturbance generated by the spark source to the far focal point. Experimental studies using both contact and non-contact sensors indicate that the source has the capability to excite the Rayleigh surface wave and impact-echo mode (S1-zero-group-velocity Lamb mode) in a 250 mm thick concrete slab and to enable fully air-coupled testing of concrete specimens.

Predicting the Exhaust Emissions of a Spark Ignition Engine Using Adaptive Neuro-Fuzzy Inference System

Abstract: This paper presents a fuzzy logic-based prediction method to reveal the performance and emission characteristics of a single cylinder spark ignition (SI) engine, which uses different fuel mixtures (gasoline–water macro-emulsions, which contains isopropanol) Adaptive neuro-fuzzy inference system, ANFIS, was used to determine some characteristic parameters due to the combustion, such as exhaust emissions (CO, CO2, HCs) Experimental data such as engine power, torque, engine speed, brake mean effective pressure, brake specific fuel consumption were used as training and checking inputs for the ANFIS model to provide a predictive algorithm The main purpose of this study is to provide a reliable model that can reveal different performance characteristics, which can be obtained from various gasoline–water macro-emulsions and doing this by the elimination of new experiments The preliminary results show that an acceptable ANFIS model can also be used in experimental design procedures, by providing quick data handling and the results

Actuator control for a new hybrid electromagnetic valvetrain in spark ignition engines

Abstract: Valve timings of a spark ignition engine have been limited to a narrow range of alternatives within the constraints of the fuel economy, the emissions and the dynamic performance. Variable valve ti...

Multi-spark and continuous spark ignition module, system, and method

Abstract: An ignition module, ignition system, and method for providing and generating at least two sparks in each cylinder in a single combustion cycle for RPMs over 3,000. The ignition module, system, and method is configured to detect misfires in a spark plug and take measures to alert a user of such misfires and cause an additional spark to occur prior to the completion of the cylinder's power stroke during its combustion cycle. The ignition module, system and method provides for continuous spark at high RPMs and is configured to reduce and/or eliminate engine misfire in excess of about 3,000 RPM for four stroke engines and up to 30,000 RPM for two stroke engines.

System and method for providing spark to an engine

Abstract: An approach for supplying spark to a spark ignited engine having activated and deactivated engine cylinders is disclosed. In one example, the approach increases ignition energy supplied to spark plugs of deactivated cylinders to reduce the possibility of spark plug fouling for the deactivated cylinders while continuing to supply spark to activated cylinders.

Control device for internal combustion engine

Abstract: PROBLEM TO BE SOLVED: To suppress discharge spark blowing-off (a phenomenon that the discharge spark of an ignition plug cannot be maintained longer), when air flow intensification is performed during execution of an EGR controlSOLUTION: The air flow intensification intensifying the air flow is performed during the execution of the EGR control in which EGR gas is recirculated in an intake pipe 12, so that the discharge spark of the ignition plug 25 is expanded Further, a blowing-off suppression control is executed In the blowing-off suppression control, it is determined whether the blowing-off of the discharge spark will occur during the execution of the EGR control, based on air flow intensity and ignition timing of the spark plug 25 If it is determined that the blowing-off of the discharge spark will occur, a discharge current of the spark plug 25 is increased to elongate a time until the discharge current drops below a discharge-sustaining current (a time until discharge voltage drops below discharge-sustaining voltage) so that the blowing-off of the discharge spark is suppressed A period where the discharge spark is expanded by the air flow intensification is elongated to effectively improve ignition performance of air-fuel mixture

An approximate dynamic programming method for multi-input multi-output nonlinear system

Abstract: SUMMARY Approximate dynamic programming is a useful tool in solving multi-stage decision optimal control problems. In this work, we first promote the action-dependent heuristic dynamic programming method to multi-input multi-output control system by extending its action network to multi-output form. The detailed derivation is also given. We then apply this method to the fluctuation control of a spark engine idle speed. An engine idling model is set up to verify the control effect of this method. Results here show that this method requires several iterations to suppress unbalanced combustion by manipulating spark ignition timing. This method provides an alternative for a simpler multi-input multi-output approximate dynamic programming scheme. Moreover, it has a faster iteration convergence effect. The derivation of this method also has a rigorous mathematical basis. Although illustrated for engines, this control system framework should also be applicable to general multi-input multi-output nonlinear system. Copyright © 2011 John Wiley & Sons, Ltd.

Multi-objective control design for turbocharged spark ignited air system: A switching Takagi-Sugeno model approach

Abstract: This paper proposes a linear matrix inequality (LMI) approach to the multi-objective synthesis of switching Takagi-Sugeno (T-S) controller. The design objectives of the closed-loop system include the robustness performance w.r.t model uncertainties and bounded disturbance, the convergence speed, and the constraints on the control input. All objectives are derived from the Lyapunov approach and they are formulated as LMI conditions. Thus, controller design task amounts to solving a LMI optimization problem. The air system control problem of a turbocharged spark ignited (TCSI) engine is used to illustrate the validity of the proposed method..

Nonlinear air-fuel ratio predictive control of spark ignited engines

Abstract: The aim of this paper is to present a nonlinear model-based predictive control (NMPC) for air-fuel ratio (AFR) of spark ignited (SI) engines Complexity of NMPC implementation causes that the computational requirements on storage space and online computation time have to be considered in the controller design Especially a limitation on computation time makes often the optimal control law too complex to be implemented and some suboptimal solution has to be provided to meet the requirements The point of the paper is to present the controller which utilizes the simplified control problem in a warm-start technique and allows to decrease the online computational effort caused by the nonlinear predictive controller