International Journal of High Performance Systems Architecture (11 papers in press)
A Soft Error Tolerant Register File for Highly Reliable Microprocessor Design
by Nastaran Rajaei, Ramin Rajaei, Mahmoud Tabandeh
Abstract: Dealing with radiation-induced soft errors is of the main design challenges in todays nanometer design of embedded systems especially in safety critical applications. Register file is a vulnerable section of a microprocessor that needs to be protected against soft errors. This paper proposes a soft error tolerant structure for the register file of the safety-critical embedded processors. In this structure, the double modular redundancy (DMR) technique based on a new hardware implementation is employed for the normal values. Moreover, the unused bits of the registers are used to be further redundant for the used ones for the narrow-width values. We show that the proposed structure offers much more reliability improvement in comparison with the conventional techniques for protection of register files such as DMR, triple modular redundancy and error detection and correction solutions based on Hamming code.
Keywords: Double Modular Redundancy (DMR); Triple Modular Redundancy (TMR); Register File; Single Event Upset (SEU); Soft Error.
An exclusive cache replacement policy based on read priority and dynamic sliding
by Debabala Swain, Banchhanidhi Dash
Abstract: The conventional cache replacement algorithms have massive hardware cost with ambiguous logic and measurability. Increasing the cache levels do not give a better solution for the hardware complexity and performance issues in multi-core processors.Rather number of read misses on executing the complex memory intensive program can increase the execution time in multi-core processors. This paper proposes a new way of cache replacement policy Weight based Read Priority Replacement (WRPR), which works on the read priority of a cache line in all levels of cache. By making it read prior instead of write, the cache lines with more read access are highly weighted. During replacement, the cache eviction is done from dynamic logical cache regions based on its weight. The algorithm performance is tested using multi-core cache simulator with different benchmark workloads in the SMPCache simulator. The proposed replacement policy can work on any exclusive cache level in a multi-level hierarchy. It shows an improved performance from the hit rate context.
Keywords: Exclusive cache; read priority; dynamic sliding; multi-core; multi-level hierarchy; WRPR.
Runtime Power-Aware Energy Saving Scheme for Parallel Applications
by Vaibhav Sundriyal, Masha Sosonkina
Abstract: Energy consumption has become a major design constraint in modern computing
systems. With the advent of petaflops architectures, power-efficient software
stacks have become imperative for scalability. Modern processors provide techniques, such as dynamic voltage and frequency scaling (DVFS),
to improve energy efficiency on-the-fly. Without careful application, however, DVFS and throttling may cause significant performance loss due to the system overhead. Typically, these techniques are used by constraining a priori the application performance loss, under which the energy savings are sought. This paper discusses potential drawbacks of such usage and proposes an energy-saving scheme that takes into account the instantaneous processor power consumption as presented by
the ``running average power limit'' (RAPL) technology from Intel. Thus, the need for the user to predefine a performance loss tolerance is avoided.
Experiments, performed on NAS parallel benchmarks and large-scale linear
system solvers from the pARMS package, show that the proposed
scheme saves more energy than the approaches based on the predefined performance
Keywords: Dynamic Voltage and Frequency Scaling (DVFS);Energy Modeling;Workload Modeling; Intel RAPL; NAS Parallel Benchmarks.
Research and application of an uncertain genetic neural network in landslide hazard prediction
by Yimin Mao, Jiawei Wang, Xinrong Lu, Dinghui Mao, Xiaodong Gao, Tian Liang
Abstract: Due to difficulties in obtaining and effectively processing rainfall and other uncertain factors in landslide hazard prediction, as well as the existence of local minima and the slow training speed of the standard back-propagation algorithm, a prediction method based on an uncertain genetic neural network in order to improve the hazard prediction accuracy has been proposed. The method is founded on an optimized genetic algorithm and the back-propagation neural network classification algorithm. Briefly, combining the prediction theory of landslide disaster with rainfall and other uncertainties associated with landslides, we propose the concept of separation of uncertain data, elaborate the processing methods of uncertain property data, and build the uncertain genetic neural network and a landslide hazard prediction model. The experiment conducted in the Baota district of Yan'an showed that the effective and overall accuracies of the method are 92.1% and 86.7%, respectively, and prove the feasibility of an uncertainty genetic neural network algorithm in landslide hazard prediction.
Keywords: Uncertain data; Landslide; Genetic algorithm; Back-propagation neural network; Hazard prediction.
Special Issue on: Internet of Things Principles, Methodologies, and Applications
A New Ontology Ranking Method with OntoDUIA for Ontology Retrieval System
by Jianghua Li, Boyu Li
Abstract: Ontology is being widely used for data integration and knowledge discovery in the field of data engineering. Ontology ranking is one of the important function of ontology search engine, which ranks the searched ontologies in a reasonable order. A good ranking method can help a user acquiring the satisfied ontology efficiently from a considerable amount of search results. However, existing methods in literatures are unable to rank ontologies to well meet users ranking demand due to their inherent shortcomings. In this paper, a novel ontology ranking method OntoDUIA is proposed and presented. It evaluates user-query related ontologies and ranks them based on a set of metrics of query relevance, usability, instance distribution and authority of ontology. To evaluate OntoDUIA, a series of experiments are conducted to compare the performance among OntoDUIA and some existing ontology ranking methods as well as human experts. Experimental results show that OntoDUIA can effectively meet users ontology ranking demand, and it achieves stable and reliable ranking results. Finally, OntoDUIA can also be applied to ontology retrieval system.
Keywords: Ontology ranking; Relevance metrics; Usability quality metrics; Instance metrics.
Cloud Computing Resources Scheduling Optimization based on Improved Bat Algorithm via Wavelet Perturbations
by Yan Zhang, Zhaobin Liu, Fahong Yu
Abstract: Efficient scheduling of computing resources is a fundamental issue that the cloud computing needs to solve, which involves highly challenging load-balance of multiple computing resources. In order to achieve optimal balance between the executing speed, the average response time and the overall system utilization during cloud computings resource allocation, a cloud computing resources scheduling optimization algorithm is proposed based on Wavelet-Perturbation-based Bat algorithm (WPBA). The algorithm first employs wavelet perturbation to enhance Bat algorithms performance followed by population-entropy-guided substitution to control diversity and improve the converging speed and accuracy. Then it adopts WPBA to achieve resources scheduling optimization of the cloud computing. The experiment shows that using WPBA has significantly improved the overall performance of the algorithm and has also remarkably optimized the resource scheduling capability of cloud computing and heightened the overall resource utilization.
Keywords: Cloud computing; Resource dispatch; Swarm intelligence; Improved bat algorithm.
Smart Generic Cabling and Network Planning in Office Network
by Banghai Xu
Abstract: Smart city intelligently responses to various requirements including environmental protection, public security, urban services, commercial and industrial activities etc. Smart city deals with lots of key information in the whole city by utilizing modern technologies like communication technology. Enterprise network is an enormous and complex system belonging to smart city. It not only provides basic platforms of office automation but also various application services, where information can be delivered to other systems. In this paper, we design a generic cabling according to the requirement of enterprise networks. Our generic cabling is well extensible and compatible. In addition, we also present the network planning for enterprise network such that it provides good service for the whole enterprise.
Keywords: Generic Cabling; Network Planning; Smart City; Enterprise Network.
SparkCUDE: a Spark-based differential evolution for large-scale global optimization
by Hu Peng, Changshou Deng, Shunxu Peng
Abstract: Differential evolution (DE) is one of the efficient evolutionary algorithms over larger-scale global optimization problems. Recently, the new cloud computing models (such as Spark) have drawn attentions to deal with larger-scale global optimization problems. Spark provides effective support for iterative algorithms. However, we have noted that simultaneous combination of the excellent DE variant and the improved Spark computing model to enhance the optimization performance and reduce the computation times has not exploited. In this paper, we propose a Spark-based DE algorithm for larger-scale global optimization problems, called SparkCUDE, in which the Spark computation model with ring topology is introduced and the CUDE algorithm is employed as the internal optimizer. The original CUDE was proposed in our previous work, in which uniform local search enhances exploitation ability and the commensal learning is proposed to adaptively select optimal mutation strategy and parameter setting simultaneously under the same criteria. Experimental studies are conducted on the benchmark functions of CEC2010 on large-scale global optimization. Comprehensive experiments demonstrate the effectiveness and efficiency of the proposed approach.
Keywords: Differential evolution; Large-Scale global optimization; Spark; Migration strategy.
Invulnerability Analysis in Intelligent Transportation System
by Fujun Han
Abstract: In recent years, the amount of population and cars increase dramatically due to the rapid development of economy. Therefore traffic congestion becomes an urgent problem in urban transport system. Intelligent transportation systems try to solve these traffic congestion problems by utilizing technologies like information, sensing, controlling and computing etc. This paper constructs a complex network of transportation system, including 73 traffic bus routes and 1003 stations. We analyze the features of public transport system towards the view of complex network theory. The results show that our network model has most features of complex networks. Finally, we compare the robustness of transportation system under fixed attacks and random attacks. Simulation results show that it has better robustness under random attacks.
Keywords: urban public transport system; complex networks; invulnerability.
Special Issue on: Data Streams Mining and Processing Methodologies, Architectures, and Applications
Analysis of Physico-chemical Variables and their Influence on Water Quality of the Bogota River using Data Mining
by Jairo Rojas, Julian Forero, Paulo Gaona, Carlos Montengro Marin, Ruben Gonzalez Crespo
Abstract: Variation of the flow rate and the concentration of different elements within a flow of water in a river must be important factors for the discovery of patterns of behavior and predictive in terms of space and time models. Based on this, this article presents data analysis carried out based on a historical compendium of measurements on the Bogota River between the years 2008-2015, given in the results of the campaigns of monitoring provided by the Regional Autonomous Corporation CAR, through the scan tool Weka and data the use of J48 algorithm for the generation of decision trees in order to establish the influence of the physical and chemical variables in the water quality of this source, within a process of identification and interpretation at the environmental level of these factors.
Keywords: Water Quality Indicator; WQI; Data Analysis;.
Research on data mining technology for the connotation and measurement of uncertainty for reassembly dimensions
by Conghu Liu, Kang He
Abstract: The uncertainty of remanufactured parts is a key factor in the stability of remanufacturing systems. Therefore, the purpose of this paper is to identify these uncertainties and measure them to improve the optimization management level of remanufacturing production process. Contrasting the ideal dimensional accuracy, manufactured dimensional accuracy and remanufactured dimensional accuracy, we analyses connotation of uncertainty for reassembly dimensions. Then, we constructs the uncertainty measurement model for reassembly dimensions to realize quantitative measurement by entropy. So the coupling mechanism of uncertainty for reassembly dimensions is studied, and the corollary is in conformity with the reality. It can use data mining technology to optimize remanufacturing process management. Finally, the feasibility and effectiveness of the model are verified in grading selection of remanufacturing enterprise parts. This research provides support for the uncertain optimization decision for lean remanufacturing from both theoretical and practical aspects by uncertain data mining techniques.
Keywords: remanufacturing; data mining; uncertainty; entropy.