확장된 GSPN 모형과 그 응용

장태무,전서현,유찬형 東國大學校 1989 論文集 Vol.28 No.-

In the present paper, a new concept called conditional are is added to tranditional GSPN(Generalized Stochastic Petri net) of which effectiveness is proved through an application to an instruction pipeline of general-purpose CPU. The problem in applying Petri nets including GSPN to performance evaluation of computer systems is that the number of nodes of reachability graph increases exponentially with the number of places in the model. Conditional ares proposed in this paper are shown to be a good tool not only for reducing the number of places in the model, but making the task of model construction easy.

네트워크 통신량을 줄인 이 단계 요청 해결 프로토콜

장태무,진종욱 東國大學校 1998 東國論叢 Vol.37 No.-

The network traffic under classical client and server systems can be reduced by caching files or parts of a file on the side of both server and clients. But owing to consistency management message traffics, the network performance is dependent on the read/write ratio. In this paper, a two phase request resolution protocol that divides a reques into pipelined phases. (request and resolution phase) is proposed. Because client directly uses the cache of resource's owner, this protocol reduce redundant messages even if the number of write clients are large. Additionally this protocol can lessen the load of the server like communicaton agents do. The protocol of this paper is evaluated by a simulation method. The performance under various workloads is measured by running the simulator that is based on randomly generated request patterns and TCP/IP LAN environments. The result shows the proposed protocol can reduce the average request completion time by about 10 to 30% comparing to the conventional Sprite file system cache management scheme.

웹 서버 시스템에서의 자원 정보를 이용한 효율적인 부하분산 기법

장태무,명원식,한준탁,Chang Tae-Mu,Myung Won-Shig,Han Jun-Tak 한국정보처리학회 2005 정보처리학회논문지 A Vol.12 No.2

웹을 사용하는 사람들의 기하급수적으로 증가하면서 확장이 용이하며 신뢰도가 높은 웹 서버가 절실히 요구된다. 사용자의 급증은 과중한 전송량과 시스템의 부하문제를 야기 시켰으며 이를 해결하기 위한 방안으로 클러스터 시스템이 연구되고 있다. 기존의 클러스터 시스템에서는 웹 서버 간 부하가 균등하더라도 멀티미디어나 CGI 둥으로 요청 데이터 크기가 크면 특정 웹 서버의 부하와 응답 시간이 증가되는 경향이 있다. 본 논문에서는 웹 서버들이 각각 다른 컨텐츠를 갖고 CPU, 메모리 및 디스크 사용율 등의 웹 서버의 자원 정보를 이용하여 효율적으로 부하를 분산하는 기법을 제안한다. 각각 서로 다른 컨텐츠를 가지고 있는 엘 서버들은 컨텐츠들에 대한 수정, 삭제, 추가 둥 자원 정보 변경으로 인하여 깨어질 수도 있는 자원 정보 일관성을 유지하기 위해 네트워크 파일 시스템에 연결되어 운영된다. 컨텐츠를 나누어 저장함으로써 생길 수 있는 각 컨텐츠 그룹 간의 부하의 불균형의 문제는 웹 서버에 대한 재설정으로 해결하였다. 성능 실험을 통해 기존의 RR방식과 LC방식보다 제안한 기법이 최대 $50\%$의 처리율과 응답시간 향상을 보여주었다. The exponential growth of Web users requires the web serves with high expandability and reliability. It leads to the excessive transmission traffic and system overload problems. To solve these problems, cluster systems are widely studied. In conventional cluster systems, when the request size is large owing to such types as multimedia and CGI, the particular server load and response time tend to increase even if the overall loads are distributed evenly. In this paper, a cluster system is proposed where each Web server in the system has different contents and loads are distributed efficiently using the Web server resource information such as CPU, memory and disk utilization. Web servers having different contents are mutually connected and managed with a network file system to maintain information consistency required to support resource information updates, deletions, and additions. Load unbalance among contents group owing to distribution of contents can be alleviated by reassignment of Web servers. Using a simulation method, we showed that our method shows up to $50\%$ about average throughput and processing time improvement comparing to systems using each LC method and RR method.

[기획연재]우리 대학의 학부제(2) - 동국대학교 정보산업과학대학 정보산업학부

장태무 한국정보과학회 1997 정보과학회지 Vol.15 No.2

비소멸성 램을 사용한 고성능 RAID5 디스크 캐쉬 제어기

장태무,이정민 東國大學校 1999 東國論叢 Vol.38 No.-

For the last ten years, the performance of processors and memory systems has been improved dramatically. but the performance of disk systems has been improved slowly. To achieve the disk system speed-up, RAID techniques that stores data over multiple inexpensive disks are introduced. Among them, RAID 5 is known to perform better under various environments, RAID 5, however, suffers from "small-write problem." Thus the performance under OLTP environments, RAID 5 shows poor performance. To solve the problem, this paper propose a new technique, called PROC(Parity Group On Cache). When RAID 5 reads a data block, this technique caches the corresponding other data blocks and a parity block at the same time. This technique is evaluated with a simulation method. And simulation results show that PROC can reduce the average disk access time by about 30% on the average for various disk access patterns.

패리티 디틀러스터링 RAID 시스템에서의 성능 개선 방안

장태무,Chang, Tae-Mu 한국정보처리학회 2000 정보처리논문지 Vol.7 No.2

RAID는 고 병렬성과 고 가용성을 목표로 하는 대용량 저장 매체를 구축하는 방법이다. 패리티 디클러스터링을 이용한 RAID는 특히 고장이 발생한 경우에도 성능의 저하를 최소화하여 고 결함 허용도 및 가용성을 높일 수 있는 저장 장치를 구축할 수 있는 기법으로 널리 연구되어 왔다. 본 논문에서는 이러한 패리티 디클러스터링을 사용한 RAID에 스페어 유닛을 분산시킨 새로운 구성을 제안하고, 특히 이러한 분산 스페어링이 고장이 없는 정상 상태에서도 성능 개선에 유용함을 보인다. 본 논문에서 제안된 방법의 유효성은 시뮬레이션 방식으로 입증하였으며, 전반적으로 정상상태의 성능을 5-15% 정도 높일 수 있음을 알 수 있다. RAID systems have been used as a mass storage system with high parallelism and availability. Especially RAID systems with parity declustering are widely studied as a technique to provide high fault tolerancy and availability by reducing performance degradation in case of disk fuilures. In this paper, a new organization of parity declustering with distributed spare units is proposed. And in normal mode where there are no failures, it is shown that these organization can improve the performance of RAID systems. By simulation methods, it is proved that the performance of RAID system in normal mode is improved by 5 to 15%.

자체 구성 기능을 가진 디스크 배열 제어기

장태무,강봉수 동국대학교 산업기술대학원 1993 산업기술논문집 Vol.1 No.-

Disk array systems such as RAID(Redundant Array of Inexpensive Disks) allow higher transfer rate and reliablility than conventional disk systems. But we cannot expect improvements on disk seek time with these system. To solve this, we propose SODA(Self-Organizing Disk Array) system that stands on the basis of RAID, and a new parity block update method for our system. With indirect address table and freemap that reflect the disk usage state, we can reduce disk seek time considerably (expecially for writes that can be performed on the nearest free block from the head), which is shown by the simulation results.

계층버스 다중처리기 시스템에서의 효율적인 캐쉬 일관성 프로토콜

장태무,박영호 동국대학교 대학원 1992 연구논집 Vol.22 No.

계층버스 다중처리기 시스템에서의 효율적인 캐쉬 일관성 프로토콜

장태무,박영호 동국대학교 대학원 1992 大學院硏究論集-東國大學校 大學院 Vol.22 No.-

We present an efficient cache coherence protocol for shared-memory multiprocessor system employing a hierarchical bus system. The protocol a modified version of Goodman's write-once cache coherent protocol, allows the existence of multiple shared blocks in a cluster but reduces contentions on bus and memory units arising from references to upper levels. Results of simulation based on multi-stage queueing network model, show that our protocol is an improvement on the existing version of writee-once in terms of throughputs and average response times measured with varying conditions

피기백 방법을 사용한 고성능 디스크 캐쉬

장태무,홍병기 東國大學校 1991 東國論叢 Vol.30 No.-

Recently the slow access time of disk storage system becomes a major bottle-neck in computer system performance improvement. Disk caching, based on the observation of locality of disk access patterns, is one of effective solutions to this problem. In this paper, a high performance disk cache, efficient not only for read accesses but for write accesses, is proposed. To accomplish this, we adds piggyback scheme to existing disk caching algorithm, that is prefetching and ordered write backs. we shows the effectiveness of our disk caching algorithm using a simulation method.