• The Transactions of the Korea Information Processing Society
• /
• v.7 no.2
• /
• pp.426-436
• /
• 2000

Given a linear arrangement of n base stations which generate multiple types of traffic among themselves, we consider the problem of finding a set of disjoint clusters to cover n base statons so that a cluster is assigned a location server. Our goal is to minimize the total communication cost for the entire network where the cost of intra-cluster communication is usually lower than that of intercluster communication for each type of traffic. The optimization problem is transformed into an equivavalent problem using the concept of relative cost, which generates the difference of communication costs between intracluster and intercluster communications. Using the relative cost matrix, an efficient algorithm of O($mm^2$), where m is the number of clusters in a partition, is designed by dynamic programming. The algorithm also finds all thevalid partitions in the same polynomial time, given the size constraint on a cluster, and the total allowable communication cost for the entire network.

• Journal of KIISE:Software and Applications
• /
• v.31 no.12
• /
• pp.1591-1601
• /
• 2004

Clustering, a technique for the analysis of the genes, organizes the patterns into groups by the similarity of the dataset and has been used for identifying the functions of the genes in the cluster or analyzing the functions of unknown gones. Since the genes usually belong to multiple functional families, fuzzy clustering methods are more appropriate than the conventional hard clustering methods which assign a sample to a group. In this paper, a Bayesian validation method is proposed to evaluate the fuzzy partitions effectively. Bayesian validation method is a probability-based approach, selecting a fuzzy partition with the largest posterior probability given the dataset. At first, the proposed Bayesian validation method is compared to the 4 representative conventional fuzzy cluster validity measures in 4 well-known datasets where foray c-means algorithm is used. Then, we have analyzed the results of Saccharomyces cell cycle expression data evaluated by the proposed method.

• Journal of KIISE:Databases
• /
• v.31 no.3
• /
• pp.320-330
• /
• 2004

Query languages for XML use paths in a data graph to represent queries. Actually, paths in a data graph are used as a basic constructor of an XML query. User can write more expressive Queries by using Patterns (e.g. regular expressions) for paths. There are many identical paths in a data graph because of the feature of semi-structured data. Current researches for indexing XML utilize identical paths in a data graph, but such an index can grow larger than source data graph and cannot guarantee efficient access path. In this paper we propose a partitioning technique that can partition all the paths in a data graph. We develop an index graph that can find appropriate partitions for a path query efficiently. The size of our index graph can be adjusted regardless of the source data. So, we can significantly improve the cost for index graph traversals. In the performance study, we show our index much faster than other graph based indexes.

• The KIPS Transactions:PartC
• /
• v.9C no.5
• /
• pp.709-718
• /
• 2002

Cluster based VOD systems require elaborate load balancing and buffer management techniques in order to ensure real-time display for multiuser concurrently. In this paper, we propose a new load balancing technique based on the dynamic buffer partitioning in cluster based VOD servers. The proposed technique evenly distribute the user requests into each service node according to its available buffer capacity and disk access rate. In each node, the dynamic buffer partitioning technique dynamically partitions the buffer to minimize the average waiting time for the requests that access the same continuous media. The simulation results show that our proposed technique decreases the average waiting time by evenly distributing the user requests compared with the exiting techniques and then increases the throughput in each node. Particularly under the overloaded condition in the cluster server, the simulation probes that the performance of the proposed technique is better two times than the Generalized Interval Caching based technique.

• KIISE Transactions on Computing Practices
• /
• v.21 no.5
• /
• pp.387-392
• /
• 2015

MapReduce is a framework used to process large data sets in parallel on a large cluster. A group-by query is a query that partitions the input data into groups based on the values of the specified attributes, and then evaluates the value of the specified aggregate function for each group. In this paper, we propose an efficient method for processing multiple group-by queries using MapReduce. Instead of computing each group-by query independently, the proposed method computes multiple group-by queries in stages with one or more MapReduce jobs in order to reduce the total execution cost. We compared the performance of this method with the performance of a less sophisticated method that computes each group-by query independently. This comparison showed that the proposed method offers better performance in terms of execution time.

• KIISE Transactions on Computing Practices
• /
• v.21 no.4
• /
• pp.315-319
• /
• 2015

This paper addresses the design and implementation of an Android application for real-time precise motion control. To provide stable real-time performance, we implemented the application in two parts: Android service in the form of a daemon process, which periodically transfers a set of position commands for all motors through a real-time fieldbus, and Android UI application, which generates and delivers the set of position commands to the Android service. To support such a real-time motion control application, we use multi-core partitioning, which partitions the processor cores into a real-time partition to be used by the real-time motion control service and a non-real-time partition to be used by the Android application, and set up a shared buffer between them for communication. Our experiments show that we can obtain a motion control period of 2 ms with 99% task activation jitters less than ${\pm}55{\mu}s$ for a configuration where each of the four threads controls two motors in a group.

• Journal of KIISE:Computer Systems and Theory
• /
• v.28 no.11
• /
• pp.549-560
• /
• 2001

Cluster system provides attractive scalability in terms of compution power and memory size. With the advances in high speed computer network technology, cluster systems are becoming increasingly competitive compared to expensive MPPs (massively parallel processors). Load balancing is very important issue since an inappropriate scheduling of tasks cannot exploit the true potential of the system and can offset the gain from parallelization. In parallel processing program, it is difficult to predict the load of each task before running the program. Furthermore, tasks are interdependent each other in many ways. The dynamic load balancing algorithm, which evaluates each processor's load in runtime, partitions each task into the appropriate granularity and assigns them to processors in proportion to their performance in cluster systems. However, if the communication cost between processing nodes is expensive, it is not efficient for all nodes to attend load balancing process. In this paper, we restrict a processor that attend load balancing by the communication cost and the deviation of its load from the average. We simulate various models of the cluster system with parameters such as communication cost, node number, and range of workload value to compare existing load balancing methods with the proposed dynamic algorithms.

• Annals of Laboratory Medicine
• /
• v.38 no.6
• /
• pp.503-511
• /
• 2018

Background: Sampling a healthy reference population to generate reference intervals (RIs) for complete blood count (CBC) parameters is not common for pediatric and geriatric ages. We established age- and sex-specific RIs for CBC parameters across pediatric, adult, and geriatric ages using secondary data, evaluating patterns of changes in CBC parameters. Methods: The reference population comprised 804,623 health examinees (66,611 aged 3-17 years; 564,280 aged 18-59 years; 173,732 aged 60-99 years), and, we excluded 22,766 examinees after outlier testing. The CBC parameters (red blood cell [RBC], white blood cell [WBC], and platelet parameters) from 781,857 examinees were studied. We determined statistically significant partitions of age and sex, and calculated RIs according to the CLSI C28-A3 guidelines. Results: RBC parameters increased with age until adulthood and decreased with age in males, but increased before puberty and then decreased with age in females. WBC and platelet counts were the highest in early childhood and decreased with age. Sex differences in each age group were noted: WBC count was higher in males than in females during adulthood, but platelet count was higher in females than in males from puberty onwards (P <0.001). Neutrophil count was the lowest in early childhood and increased with age. Lymphocyte count decreased with age after peaking in early childhood. Eosinophil count was the highest in childhood and higher in males than in females. Monocyte count was higher in males than in females (P <0.001). Conclusions: We provide comprehensive age- and sex-specific RIs for CBC parameters, which show dynamic changes with both age and sex.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.11
• /
• pp.51-59
• /
• 2019

Due to the exponential growth of access information on the web, the need for predicting web users' next access has increased. Various models such as markov models, deep neural networks, support vector machines, and fuzzy inference models were proposed to handle web access prediction. For deep learning based on neural network models, training time on large-scale web usage data is very huge. To address this problem, deep neural network models are trained on cluster of computers in parallel. In this paper, we investigated impact of several important spark parameters related to data partitions, shuffling, compression, and locality (basic spark parameters) for training Multi-Layer Perceptron model on Spark standalone cluster. Then based on the investigation, we tuned basic spark parameters for training Multi-Layer Perceptron model and used it for tuning Spark when training Multi-Layer Perceptron model for web access prediction. Through experiments, we showed the accuracy of web access prediction based on our proposed web access prediction model. In addition, we also showed performance improvement in training time based on our spark basic parameters tuning for training Multi-Layer Perceptron model over default spark parameters configuration.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.19 no.4
• /
• pp.65-69
• /
• 2019

This paper presents an optimized adoption of NVM for the storage system considering application characteristics. To do so, we first characterize the storage access patterns for different application types, and make two prominent observations that can be exploited in allocating NVM storage efficiently. The first observation is that a bulk of I/O does not happen on a single storage partition, but it is varied significantly for different application categories. Our second observation is that there exists a large proportion of single accessing in storage data. Based on these observations, we show that maximizing the storage performance with NVM is not obtained by fixing it as a specific storage partition but by allocating it adaptively for different applications. Specifically, for graph, database, and web applications, using NVM as a swap, a journal, and a file system partitions, respectively, performs well.