• Journal of the Korea Society of Computer and Information
• /
• v.19 no.7
• /
• pp.113-130
• /
• 2014

In this paper, we propose BLOCS algorithm to find sequence of data that should be saved in cache device of hybrid storage system which uses SSD as a cache device. BLOCS algorithm which uses a sequence pattern mining scheme, creates a set of frequently requested sectors with respect to requested order of sectors. To compare the performance of the proposed scheme, we introduce Distance (DIST) based scheme, Request Frequency (FREQ) based scheme, and Frequency times Size (F-S) based scheme. We measure the hit ratio and I/O latency of different caching schemes using hybrid storage caching simulator. We acquired booting workload along with ten scenarios of launching applications and use the workloads as input to the cache simulator. After experiment with booting workload, we find that BLOCS scheme gives hit ratio of 61% which is about 15% higher than the least performing DIST scheme.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.11
• /
• pp.1031-1040
• /
• 2010

Hash join is one of the core algorithms in databases management systems. If a hash join cannot complete in one-pass because the available memory is insufficient (i.e., hash table overflow), however, it may incur a few sequential writes and excessive random reads. With harddisk as the tempoary storage for hash joins, the I/O time would be dominated by slow random reads in its probing phase. Meanwhile, flash memory based SSDs (flash SSDs) are becoming popular, and we will witness in the foreseeable future that flash SSDs replace harddisks in enterprise databases. In contrast to harddisk, flash SSD without any mechanical component has fast latency in random reads, and thus it can boost hash join performance. In this paper, we investigate several important and practical issues when flash SSD is used as tempoary storage for hash join. First, we reveal the va patterns of hash join in detail and explain why flash SSD can outperform harddisk by more than an order of magnitude. Second, we present and analyze the impact of cluster size (i.e., va unit in hash join) on performance. Finally, we emperically demonstrate that, while a commerical query optimizer is error-prone in predicting the execution time with harddisk as temporary storage, it can precisely estimate the execution time with flash SSD. In summary, we show that, when used as temporary storage for hash join, flash SSD will provide more reliable cost estimation as well as fast performance.

• Proceedings of the IEEK Conference
• /
• 2005.11a
• /
• pp.707-710
• /
• 2005

In this paper, we propose a new communication network for on-chip communication. The network is based on a direct sequence code division multiple access (DS-CDMA) technique. The new communication network is suitable for a parallel processing system and also drastically reduces the I/O pin count. Our network architecture is mainly divided into a CDMA-based network interface (CNI), a communication channel, a synchronizer. The network includes a reverse communication channel for reducing latency. The network decouples computation task from communication task by the CNI. An extreme truncation is considered to simplify the communication link. For the scalability of the network, we use a PN-code reuse method and a hierarchical structure. The network elements have a modular architecture. The communication network is done using fully synthesizable Verilog HDL to enhance the portability between process technologies.

• Proceedings of the KOSOMBE Conference
• /
• v.1995 no.11
• /
• pp.265-268
• /
• 1995

Signal averaging technique to improve signal-to-noise ratio has widely been used in various fields, especially in electrophysiology. Estimation of the EP(evoked potential) signal using the conventional averaging method fails to correctly reconstruct the original signal under EEG(electroencephalogram) noise especial]y when the latency times of the evoked potential are not identical. Therefore, a technique based on the bispectrum averaging was proposed for recovering signal waveform from a set o noisy signals with variable signal dalay. In this paper an improved bispectrum estimation technique of the RP signal is proposed using a confidence thresholding of the EP signal in frequency domain in which energy distribution of the EP signal is usually not uniform. The suggested technique is coupled with the conventional bispectrum estimation technique such as least square method and recursive method. Some results with simulated data and real EP signal are shown.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.7 no.3
• /
• pp.129-134
• /
• 2012

The objective of cache is to reduce I/O access of physical storage device so that user accesses their data faster. Traditionally, the most important metric to measure the performance of cache is hitratio. Thus, when the cache maintains hitratio high, it is regarded as a good cache replacement policy. However, the cache miss latency is different when the storages are heterogeneous. Though the cache hitratio is high, if the cache often misses with low performance disk, then the user experiences low performance. To address this problem we proposed eviction cost estimation based cache management. In our result, the eviction cost estimation based cache management has 10~30% throughput improvement compared with LRU cache management.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04a
• /
• pp.79-81
• /
• 2003

스트라이핑(striping) 기법은 여러 개의 디스크에 데이터를 병렬적으로 분산하여 저장함으로써 데이터 접근 속도를 높이기 위하여 제안되었다. 그러나 서로 다른 데이터를 요청하는 다수의 사용자 요구 환경에서 스트라이핑 기법은 seektime과 rotational latency의 증가로 인해. 오히려 I/O 성능이 저하될 수 있다. 본 논문에서는 서로 다른 데이터 요구율을 갖는 다중 데이터베이스 액세스 환경에서 스트라이핑 폭(width)에 의한 입출력 성능을 분석하고, 사용자 요구 형태에 따라 스트라이핑 디스크 그룹과 개별 디스크 그룹으로 디스크 풀(pool)을 나누어 처리하는 하이브리드 스트라이핑 방식을 제안한다. 제안 방식의 성능 향상은 시뮬레이션에 의하여 검증하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.11
• /
• pp.5252-5268
• /
• 2018

To support a large-scale Hadoop cluster, Hadoop heartbeat messages are designed to deliver the significant messages, including task scheduling and completion messages, via piggybacking to reduce the number of messages received by the NameNode. Although Hadoop is designed and optimized for high-throughput computing via batch processing, the real-time processing of large amounts of data in Hadoop is increasingly important. This paper evaluates Hadoop's performance and costs when the heartbeat period is controlled to support latency sensitive applications. Through an empirical study based on Hadoop 2.0 (YARN) architecture, we improve Hadoop's I/O performance as well as application performance by up to 13 percent compared to the default configuration. We offer a guideline that predicts the performance, costs and limitations of the total system by controlling the heartbeat period using simple equations. We show that Hive performance can be improved by tuning Hadoop's heartbeat periods through extensive experiments.

• Natural Product Sciences
• /
• v.22 no.4
• /
• pp.263-269
• /
• 2016

Rhynchophylline (RP) is a major tetracyclic oxindole alkaloid of Uncariae Ramulus et Uncus which has been used to treat hypertension, seizures, pain and anxiety in the oriental countries. A recent report revealed that RP attenuated ischemia-induced neuronal damage and kainite-induced convulsions in animals. This study was performed to investigate whether RP enhances pentobarbital-induced sleep behaviors and modulates sleep architecture in mice. Locomotor activity was significantly inhibited by RP at 0.25 and 0.5 mg/kg, similar to 2 mg/kg diazepam (a benzodiazepine agonist) in mice. RP shortened sleep latency and increased total sleep time in a dose-dependent manner when administrated with pentobarbital (42 mg/kg, i.p.). RP also increased the number of sleeping mice and total sleep time by concomitant administration with the sub-hypnotic dosage of pentobarbital (28 mg/kg, i.p.). On the other hand, RP (0.25 mg/kg, p.o.) itself significantly inhibited sleep-wake cycles, prolonged total sleep time, and rapid eye movement in rats. In addition, RP also increased chloride influx in the primary cultured hypothalamic neuronal cells. In addition, we found that glutamic acid decarboxylase ($GAD_{65/67}$) was activated by RP. In conclusion, RP augments pentobarbital-induced sleeping behaviors, and can be a candidate for treating insomnia.

• The Journal of the Korea Contents Association
• /
• v.10 no.2
• /
• pp.54-62
• /
• 2010

Inverted index structures have become the most efficient data structure for high performance indexing of large text collections, especially online index maintenance, In-Place and merge-based index structures are the two main competing strategies for index construction in dynamic search environments. In the above-mentioned two strategies, a contiguity of posting information is the mainstay of design for online index maintenance and query time. Whereas with the emergence of new storage device(SSD, SCRAM), those do not consider a contiguity of posting information in the design of index structures because of its superiority such as low access latency and I/O throughput speeds. However, SSD(Solid State Drive) is not well suited for traditional inverted structures due to the poor random write throughput in practical systems. In this paper, we propose the new efficient online index structure(SPM) for SSD that significantly reduces the query time and improves the index maintenance performance.

• Journal of Korea Multimedia Society
• /
• v.8 no.4
• /
• pp.502-508
• /
• 2005

This paper proposes a dynamic hybrid patching technique that can improve the performance of VOD systems by adopting the advantages of both greedy patching and grace patching to use a VOD server's limited network-I/O bandwidth. The proposed technique uses grace patching to the requests for the videos, arrival intervals of which are smaller than the size of patching window, and uses greedy patching to the rest requests. In addition, proposed patching technique expands the latest particular patching multicast into a regular multicast for a new request. In result, the patching multicast data for the new request can be the data from the beginning to the skew point of the video and the holding time of a dispatched channel can exceedingly decrease. Simulation results confirm that the proposed technique is better than grace patching in terms of defection rate and average service latency.