http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
盧三赫,廉美鈴 弘益大學校 科學技術硏究所 2001 科學技術硏究論文集 Vol.12 No.-
We propose the DoubleQ web server that classifies incoming requests according to whether the requested document is cached or not. Requests that demand a cached document are put into the Service Queue while other requests are added to the Defer Queue. The DoubleQ web server services requests that are in the Service Queue before it services requests in the Defer Queue. This strategy is used to reduce disk accesses that have been the predominant overhead of traditional web servers. Experimental results using synthetic data show that improvements in the average user response time and the throughput of the web server may be achieved.
A Study of a Heterogeneous Architecture for Massively Parallel Processing
盧三赫 弘益大學校 科學技術硏究所 1994 科學技術硏究論文集 Vol.5 No.-
본 논문에서는 이질 프로세서, 즉, 성능에서 차이를 보이는 프로세서로 구성되고 SIMD와 MIND설계를 강점들을 조합한 초병렬 시스템을 제시한다. 이 HeMM (Heterogeneous Mixed-Mode) 시스템은 두 개의 주요 구성 요소들이 있다. 연산의 주는 초병렬로 구성되며 data parallel방식으로 연산하는 SIMD요소에서 이루어진다. 반면, MIMD요소는 작은 수 (또는 한 개)의 강력한 프로세서로 구성되며 control parallel방식으로 연산한다. 이렇게 구성된 시스템은 병렬 처리의 강점과 함께 프로그램의 변화하는 병렬성에 쉽게 대응할 수 있다는 강점을 보유한다. 이 논문은 이 시스템을 바탕으로 성능 평가를 하여 이렇한 시스템이 제시하는 성능의 향상을 분석한다.
盧三赫,張惠眞 弘益大學校 科學技術硏究所 2000 科學技術硏究論文集 Vol.11 No.-
Web caching plays an important role in Web server perfoiniance improvement. To improve the performance of Web caching, we apply the technique of delayed replacement for enhanced services. To measure the performance, we use a synthetic log created from the SURGE program and a real log obtained from the UC Berkeley Computer Science department server. We compare the hit rates and byte hit rates of the LRU replacement policy with the delayed replacement policy.
盧三赫,都仁煥 弘益大學校 科學技術硏究所 2003 科學技術硏究論文集 Vol.14 No.-
Web services are used in a various area and clients are increasing every year. Hence, the need for web servers requiring better system performance is ever increasing. To resolve this problem, we are consider scalable web cluster systems. In this paper, we develop a web cluster system based on a Layer-4 switch, implemented in the Linux kernel 2.4.20. To further assist content-aware requests distribution, the web cluster system internally includes a Layer-7 switch. In addition, a light-weight transfer protocol that replaces the TCP control connection is designed and implemented to enhance performance. Our current state of the implementation shows that the system is stable. The issue of scalability in this system is left for future work.
嚴世雄,盧三赫 弘益大學校 科學技術硏究所 1996 科學技術硏究論文集 Vol.7 No.2
I/O subsystems are becoming a serious bottleneck of system performance. Recently, we have seen a concentration of research dedicated to alleviating this problem. Research in such areas as caching , prefetching, parallel file systems, and RAID are some examples. However, these studies have mostly concentrated on single stand-alone systems. The expansion of these ideas to distributed systems have been limited. This paper describes the simulation framework needed to extend the studies on these areas. Details of the design and implementation of a simulator for file systems in a distributed environment are described.
元裕憲,金鍾勳,盧三赫 弘益大學校 科學技術硏究所 1995 科學技術硏究論文集 Vol.6 No.-
Disk arrays were proposed in the 1980`s as a way to allow parallelism between multiple disks to improve aggregate I/O performance. Today they appear in the product lines of most major computer manufactures. This paper gives a comprehensive overview of disk arrays and provides a framework in which to organize current and future work. First, this paper introduces disk arrays in general, and reviews the driving forces that have popularized disk arrays : performance and reliability. It discusses the two architectural techniques used in disk arrays, striping across multiple disks to improve performance and redundancy to improve reliability. Next, this paper describes 6 disk array architectures, namely RAID levels 0∼5. It goes on to discuss advanced research and implementation topics such as the way in which parity is distributed in a RAID, and the means of refining the basis RAID levels to improve performance. Finally, this paper describes software disk arrays which allow writing data redundantly across an array of disks in workstations distributed across the network.
金태延,張惠眞,盧三赫 弘益大學校 科學技術硏究所 1999 科學技術硏究論文集 Vol.10 No.1
In this paper, we present a study of the effect of prefetching of pages in the virtual memory subsystem of the operating system. As disk accesses are increasingly becoming a bottleneck of a computer system it is essential that this bottleneck be alleviated. Prefetching of pages in one such approach. Using an algorithm proposed for buffer caching, we modify the virtual memory subsystem of the FreeBSD operating system kernel so that prefetching may be achieved in an effective manner. To validate the effect of prefetching we run experiments using the original operating system and the modified operating system. The experiments are conducted using matrix calculations and by running a I/O intensive benchmark. Results show that the modified kernel provides more efficient execution in terms of actually reducing the number of page faults as well as the real execution time. However, the experiments also show that the bookkeeping overhead induced by the modification increases the system overhead slightly.