• Proceedings of the Korean Information Science Society Conference
• /
• 2004.04b
• /
• pp.217-219
• /
• 2004

MMDB에서 성능에 가장 근 영향을 미치는 부분은 로깅. 체크포인트, 락이다. 기존 연구는 이것들의 성능에 많은 초점을 맞추었으며, 이는 레코드 저장노드 크기에 관한 연구도 마찬가지다. 하지만 초고성능을 요하는 최신 MMDB 응용에서는 성능 조건을 충족시키기 위해 로깅, 체크포인트, 락을 포기하고 기본적인 레코드 저장구조 기능만으로 MMDB를 운용하기도 한다. 이 경우 레코드 저장구조 성능이 중요하게 된다. 이 논문에서는 실험을 통해 최적의 레코드 저장구조 성능을 보이는 레코드 저장노드 크기를 구한다. 그리고 실험 결과를 CPU 캐시 친화 관점에서 분석한다. 최종적으로는 MMDB 시스템 전체 성능의 최적화 관점에서, 실험 결과 얻어진 레코드 저장노드 크기를 검토한다.

• Journal of the Semiconductor & Display Technology
• /
• v.21 no.3
• /
• pp.1-6
• /
• 2022

Modern computer systems usually have special hardware for operations used in deep learning workload even edge computing environment. Non-volatile memories (NVMs) have been considered for alternative memory storage because they consume little static energy and occupy small area. However, there is a problem for NVMs to be directly adopted. An NVM cell has limited write endurance, so that the lifetime of NVM-based memory system is much shorter than that of conventional memory system. To overcome this problem for the deep learning system, this paper proposes a novel method to extend the lifetime based on the analysis of the deep learning workloads. If an incoming block has more than a predefined number of frequently used values, the cacheline is defined as write friendly block. During the victim selection, the cacheline has lower possibility to be chosen as victim. The experimental results show that the lifetime is increased by about 50% and energy consumption is decreased by 3% with a little performance hurt.

• Annual Conference of KIPS
• /
• 2009.11a
• /
• pp.235-236
• /
• 2009

비휘발성 나노 저장 소자는 고속의 바이트 단위 접근성과 함께 비휘발성을 동시에 갖고 있다. 이와 같은 특징은 차세대 장치로 주목 받을 만큼 오늘날의 컴퓨터 구조에 큰 변화를 줄 수 있는 잠재력을 갖고 있으며 이를 접목한 시스템적인 개발 역시 활발하게 진행되고 있다. 본 논문에서는 기존의 메인 메모리와 나노 저장 소자가 융합된 스토리지 메모리 클래스(SCM) 환경 하에서 입출력시에 원자성(Atomicity)이 보장되도록 설계, 구현된 새도우 블록 기법을 소개하고, 더불어 캐시를 사용하며 발생할 수 있는 데이터 일관성 처리의 보장을 다루었다. 또한 실제 FRAM이 장착된 하드웨어 환경 하에서 개선된 새도우 블록을 동작하여 측정한 성능 결과를 함께 제공한다.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.2
• /
• pp.237-241
• /
• 2010

Energy conservation of server systems has become important. Though storage subsystem is one of the biggest power consumers, development of energy conservation techniques is challenging problem due to striping techniques like RAID and physical characteristics of hard disks. According to our observation, the size of footprint for a day or for hours is much smaller compared to the size of whole data set. In this paper, we describe a design of a novel architecture for RAID that uses an SSD as a large cache to conserve energy by holding such a footprint. We incorporate these approaches into a real implementation of a RAID 5 system that consists of four hard disks and an SSD in a Linux environment. Our preliminary results in actual performance measurements using the cello99 and SPC traces show that energy consumption is reduced by a maximum of 14%.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.2
• /
• pp.1-8
• /
• 2011

NAND type Flash memory has performing much researches for a hard disk substitution due to its low power consumption, cheap prices and a large storage. Especially, the NAND type flash memory is using general buffer systems of a cache memory for improving overall system performance, but this has shown a tendency to emphasize in terms of data. So, our research is to design a high performance instruction NAND type flash memory structure by using a buffer system. The proposed buffer system in a NAND flash memory consists of two parts, i.e., a fully associative temporal buffer for branch instruction and a fully associative spatial buffer for spatial locality. The spatial buffer with a large fetching size turns out to be effective serial instructions, and the temporal buffer with a small fetching size can achieve effective branch instructions. According to the simulation results, we can reduce average miss ratios by around 77% and the average memory access time can achieve a similar performance compared with the 2-way, victim and fully associative buffer with two or four sizes.

• KIISE Transactions on Computing Practices
• /
• v.22 no.2
• /
• pp.95-100
• /
• 2016

In the current era of big data and cloud computing, the amount of data utilized is increasing, and various systems to process this big data rapidly are being developed. A distributed file system is often used to store the data, and glusterFS is one of popular distributed file systems. As computer technology has advanced, NAND flash SSDs (Solid State Drives), which are high performance storage devices, have become cheaper. For this reason, datacenter operators attempt to use SSDs in their systems. They also try to install glusterFS on SSDs. However, since the glusterFS is designed to use HDDs (Hard Disk Drives), when SSDs are used instead of HDDs, the performance is degraded due to structural problems. The problems include the use of I/O-cache, Read-ahead, and Write-behind Translators. By removing these features that do not fit SSDs which are advantageous for random I/O, we have achieved performance improvements, by up to 255% in the case of 4KB random reads, and by up to 50% in the case of 64KB random reads.

• The KIPS Transactions:PartA
• /
• v.14A no.4
• /
• pp.235-244
• /
• 2007

Caching and prefetching policies have been used in most of computer systems to compensate speed differences between primary memory and secondary storage devices. In this paper, we design and implement a Flash Cache Core Module(FCCM) on the YAFFS which operates on a file system layer for NAND flash memory. The FCCM is independent of the underlying kernel in order to support its stability and compatibility. Also, we implement the Dirty-Last memory replacement technique considering the characteristics of flash memory, and the waiting queue for pages to be prefetched according to page hit. The FCCM reduced the number of I/Os and the amount of prefetched pages by maximum 55%(20% on average) and maximum 55%(24% on average), respectively, comparing with caching and prefetching policies of Linux.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.4
• /
• pp.82-88
• /
• 2009

Video proxy server which is located near clients can store the frequently requested video data in storage space in order to minimize initial latency and network traffic significantly. However, due to the limited storage space in video proxy server, an appropriate video selection method is needed to store the videos which are frequently requested by users. Thus, we present a virtual caching technique to efficiently store the video in video proxy server. For this purpose, we employ a virtual memory in video poky server. If the video is requested by user, it is loaded in virtual memory first and then, delivered to the user. A video which is loaded in virtual memory is deleted or moved into the storage space of video poxy sewer depending on the request condition. In addition, virtual memory is divided into each segment area in order to store the segments efficiently and to avoid the fragmentation. The simulation results show that the proposed method performs better than other methods in terms of the block hit rate and the number of block deletion.

• The KIPS Transactions:PartA
• /
• v.13A no.4 s.101
• /
• pp.343-350
• /
• 2006

In this paper, wireless internet proxy server clusters are used for the wireless internet because their caching, distillation, and clustering functions are helpful to overcome the limitations and needs of the wireless internet. A wireless Internet proxy server cluster needs a systematic scalability, simple communication structure, cooperative caching, and serving Hot Spot requests. In our former research, we proposed the CD-A structure which can be scalable in a systematic way and has a simple communication structure but it has no cooperative caching. A hash based load balancing can be used to solve the problem, but it can not deal with Hot Spot request problem. In this paper, we proposed a shared storage based wireless internet proxy server cluster which has a systematic scalability, simple communication structure, cooperative caching, and serving Hot Spot requests. The proposed method shares one cache directory and it has advantages: advantages of the existing CD-A structure, cooperative caching, and serving Hot Spot requests. We performed experiments using 16 PCs and experimental results show high performance improvement of the proposed system compared to the existing systems in Hot Spot requests.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.4
• /
• pp.2276-2291
• /
• 2017

Data protection of removable storage devices is an important issue in information security. Unfortunately, most existing data protection mechanisms are aimed at protecting computer platform which is not suitable for ultra-low-power devices. To protect the flash disk appropriately and efficiently, we propose a trust based USB flash disk, named UTrustDisk. The data protection technologies in UTrustDisk include data authentication protocol, data confidentiality protection and data leakage prevention. Usually, the data integrity protection scheme is the bottleneck in the whole system and we accelerate it by WH universal hash function and speculative caching. The speculative caching will cache the potential hot chunks for reducing the memory bandwidth pollution. We adopt the symmetric encryption algorithm to protect data confidentiality. Before mounting the UTrustDisk, we will run a trusted virtual domain based lightweight virtual machine for preventing information leakage. Besides, we prove formally that UTrustDisk can prevent sensitive data from leaking out. Experimental results show that our scheme's average writing throughput is 44.8% higher than that of NH scheme, and 316% higher than that of SHA-1 scheme. And the success rate of speculative caching mechanism is up to 94.5% since the access pattern is usually sequential.