• Journal of Platform Technology
• /
• 제11권4호
• /
• pp.71-79
• /
• 2023

이 논문에서는 기계학습 워크로드의 특징을 분석하고 이를 기반으로 기계학습 워크로드의 성능 향상을 위한 분산 인-메모리 캐싱 기법을 제안한다. 기계학습 워크로드의 핵심은 모델 학습이며 모델 학습은 컴퓨팅 집약적 (Computation Intensive)인 작업이다. 쿠버네티스 기반 클라우드 환경에서 컴퓨팅 프레임워크와 스토리지를 분리한 구조에서 기계학습 워크로드를 수행하는 것은 자원을 효과적으로 할당할 수 있지만, 네트워크 통신을 통해 IO가 수행되야 하므로 지연이 발생할 수 있다. 이 논문에서는 이런 환경에서 수행되는 머신러닝 워크로드의 성능을 향상하기 위한 분산 인-메모리 캐싱 기법을 제안한다. 특히, 제안하는 방법은 쿠버네티스 기반의 머신러닝 파이프라인 관리 도구인 쿠브플로우를 고려하여 머신러닝 워크로드에 필요한 데이터를 분산 인-메모리 캐시에 미리 로드하는 새로운 방법을 제안한다.

• 한국시뮬레이션학회논문지
• /
• 제2권1호
• /
• pp.91-106
• /
• 1993

Load sharing in distributed systems improves systems performance. Research in the field has been focused on steady state performance for load sharing algorithms. However, transient characteristics of such algorithms may be important in a distributed system in which workload for some node is changing. Simulation is the only means to analyze such characteristics. This paper presents a simulation-based analysis of the transient characteristics of four load sharing algorithms ; forward probing, reverse probing, symmetric probing, and multi-threshold symmetric probing algorithms. Discrete event system models for execution of the algorithms in a distributed system has been developed in a SIMSCRIOT II.5 environment. Simulation results indicate that the MSYM algorithm shows the shortest response time in the transient period.

• 한국환경보건학회지
• /
• 제43권5호
• /
• pp.370-381
• /
• 2017

Objectives: The rate at which the nationwide highway is outsourcing the duties of toll gate cashiers to the public is increasing. The aim of this study is to conduct research on job stress for the rate at which the fares are collected by national highway toll gate cashiers. Methods: The population selected was a statistically significant group making up 10% of the total number of people surveyed in 2015, with a population of approximately 5,495 persons. The sample size is intended to cover the maximum sample size of 550 samples and the maximum number of samples that should be obtained. After allocating the number of samples by quota allocation, the qualitative function of Quality Function Deployment (QFD) was applied by random sampling. Results: Job stress ranked at $46.6{\pm}11.5$ points was found to be under the stress of 50% reported in other occupations. Job stress items showed higher for "job instability" compared to other types of tasks. Also, it appeared lower in "conflict in relationships." This is a result of the work characteristics of highway service operators. Conclusions: Time was found to have the highest influence on the workload (${\beta}=0.80$, p. 0.001), which calls for a reduction in time and workload. est influence on the workload (${\beta}=0.80$, p. 0.001), which requires a reduction in time and workload.

• 반도체디스플레이기술학회지
• /
• 제22권3호
• /
• pp.142-148
• /
• 2023

In this paper, we identify performance issues in executing compute kernels from PolyBench, which includes compute kernels that are the core computational units of various data-intensive workloads, such as deep learning and data-intensive applications, on Processing-in-Memory (PIM) devices. Therefore, using our in-house simulator, we measured and compared the various performance metrics of workloads based on traditional out-of-order and in-order processors with Processing-in-Memory-based systems. As a result, the PIM-based system improves performance compared to other computing models due to the short-term data reuse characteristic of computational kernels from PolyBench. However, some kernels perform poorly in PIM-based systems without a multi-layer cache hierarchy due to some kernel's long-term data reuse characteristics. Hence, our evaluation and analysis results suggest that further research should consider dynamic and workload pattern adaptive approaches to overcome performance degradation from computational kernels with long-term data reuse characteristics and hidden data locality.

• 대한치과기공학회지
• /
• 제33권4호
• /
• pp.379-390
• /
• 2011

Purpose: This study was conducted to improve the working conditions of dental technicians, through survey for working conditions related to perform one's duty, the extent of personal exposure to substances hazardous and complaint rate and factor of job stress and subjective symptom on musculoskeletal disorders. Methods: Using four types of structured questionnaires: social and demographic factor; the actual conditions of working space; working conditions; and characteristics of dental technicians, respondents filled in the questionnaires and the results were analyzed statistically. Results: The evaluation of comfort in working place found that satisfaction level on indoor noise was the lowest with 2.6 on a five-point scale and privacy space followed 2.8 point. In reasons for dissatisfaction with job, over workload ranked the highest by 25.4% and role overload had the highest by 39.5% in job stress factors. Based on the results, we could draw conclusion that dental technicians were unsatisfied with role overload and heavy workload a day. In reasons for musculoskeletal symptom prevalence, long-patient work was the highest by 24.1%. It was thought to cause dental technicians need deep procession and long-patient work due to the nature of their job. Conclusion: In conclusion, noise reduction and privacy space are required to improve job efficiency of dental technicians. In addition; we need to think of ways to increase the job satisfaction with adequate rest while deep procession and long-patient working.

• 정보처리학회논문지D
• /
• 제11D권4호
• /
• pp.747-754
• /
• 2004

데이타베이스 시스템을 효율적으로 사용하기 위해 데이타베이스 관리자는 데이타베이스 시스템의 성능 특징들을 알아야 한다 데이터베이스 시스템에서 자원 사용은 워크로드에 따라 다르게 나타난다. 본 논문의 목적은 상이할 워크로드에서 데이타베이스 튜닝에 도움이 될 수 있도록 데이타베이스 시스템의 성능 특징 분석하고 식별하는 것이다. 이를 위해, OLTP 환경을 나타내는 TPC-C 워크로드와 웹기반의 전자상거래 환경을 나타내는 TPC-W워크로드에서 14개의 성능 지표에 의해 결정되는 자원 사용 형태를 분석하고 4개(데이타 버퍼, 개인 메모리, I/O 프로세스, 공유 메모리)의 자원 할당 변경에 따른 자원 사용 형태의 변화를 분석한다. 분석에 대한 결과로서 14개 중 8개의 성능 지표는 워크로드에서 성능차이를 보이고, 데이타 버퍼 자원의 변경은 데이타베이스 시스템에 영향을 준다. 본 논문의 결과는 데이타베이스 시스템 자동 튜닝의 기초 자료로서 사용될 수 있다.

• 보건교육건강증진학회지
• /
• 제19권3호
• /
• pp.107-120
• /
• 2002

There has been a rapid industrial progress in Korea since 1962 by the success of 5-year economic development plan, and the number of industrial work has also made a rapid increase. Consequently, the management of the occupational health for the purpose of promoting the health of industrial workers and improving the working environment is badly needed in these days. Health services on industrial noisy environment have been provided only for noise-induced hearing loss management until now. But gradually, modem diseases and death have come to be related to the stress and mental health, therefore noise-induced mental disorder, like a stress became very important. Thus, this study has been carried out to analyze the relationship between workers' stress symptoms and the perceived working environment and the perceived working conditions. This study included 786 industrial workers selected from II factories in Buchun. The results were as follows: 1. For demographic characteristics, most of the workers were males(75.7%), the 20～29 years old were 33.8% and those who graduated from high school were 56.1%. The workers whose monthly income ranged from 700,000 to 1,500,000 won were 37.9% and who has a religion were 49.0%. 2. For occupational characteristics, workers who had worked 5～10 years in the factories were 35.8%. Those who felt much for them workload were 42.7% and who worked more than 8 hours a day were 73.7%. Those who were dissatisfied with their pay and job were 51.1 % and 31.2%. The workers who responded ventilation condition of their worksites were bad were 50.4% and the dissatisfied with working environment of their worksites were 43.8%. 3. For the noise exposure level in worksite, workers who were exposed to 70∼90㏈ were 37.4%, 90∼100㏈ were 25.2% and 50∼70㏈ were 18.8%. 4. Workers∼ stress symptoms were significantly related to marital status and their monthly income(P〈0.05). Workers who were single and had lower monthly income showed higher PSI(Psychiatric Symptom Index) scores than those who were married and had higher monthly income. Higher PSI scores were also significantly related to the night-work, workload, dissatisfaction with their job, and bad relationship with their bosses or co-workers. 5. The higher noise exposure level in worksite from 80㏈ was, the more severe stress symptoms including PSI subparts were reported; Anxiety, Anger, Depression, and Cognitive disorder symptom(P〈0.001). 6. According to the results of stepwise multiple regression analysis, factors affecting workers' PSI scores were noise exposure level in worksite(R2=0.150), relationship with coworkers, amount of workload, monthly income and relationship with bosses orderly and the total R2 of this 6 factors was 29.7%.

• ETRI Journal
• /
• 제36권6호
• /
• pp.1023-1031
• /
• 2014

Failure analysis is necessary to clarify the root cause of a failure, predict the next time a failure may occur, and improve the performance and reliability of a system. However, it is not an easy task to analyze and interpret failure data, especially for complex systems. Usually, these data are represented using many attributes, and sometimes they are inconsistent and ambiguous. In this paper, we present a scalable approach for the analysis and interpretation of failure data of high-performance computing systems. The approach employs rough sets theory (RST) for this task. The application of RST to a large publicly available set of failure data highlights the main attributes responsible for the root cause of a failure. In addition, it is used to analyze other failure characteristics, such as time between failures, repair times, workload running on a failed node, and failure category. Experimental results show the scalability of the presented approach and its ability to reveal dependencies among different failure characteristics.

• International journal of advanced smart convergence
• /
• 제12권4호
• /
• pp.164-170
• /
• 2023

With the recent advances in AI (artificial intelligence) and HPC (high-performance computing) technologies, deep learning is proliferated in various domains of the 4th industrial revolution. As the workload volume of deep learning increasingly grows, analyzing the memory reference characteristics becomes important. In this article, we analyze the memory reference traces of deep learning workloads in comparison with traditional workloads specially focusing on read and write operations. Based on our analysis, we observe some unique characteristics of deep learning memory references that are quite different from traditional workloads. First, when comparing instruction and data references, instruction reference accounts for a little portion in deep learning workloads. Second, when comparing read and write, write reference accounts for a majority of memory references, which is also different from traditional workloads. Third, although write references are dominant, it exhibits low reference skewness compared to traditional workloads. Specifically, the skew factor of write references is small compared to traditional workloads. We expect that the analysis performed in this article will be helpful in efficiently designing memory management systems for deep learning workloads.

• 한국식품영양과학회지
• /
• 제46권1호
• /
• pp.132-139
• /
• 2017

본 연구는 부산지역 초 중 고등학교 영양(교)사를 대상으로 직무변인(업무량, 자율성, 보상, 커뮤니티, 공정성, 가치)이 직무소진(정서적 소진, 냉소)과 직무열의(활기, 헌신, 몰입)에 미치는 영향력을 규명함으로써 영양(교)사의 직무개선에 필요한 자료를 제공하고자 시도하였다. 본 연구의 목적을 위해 부산광역시 초 중 고등학교 영양(교)사를 대상으로 설문조사를 시행하였다. 2014년 2월 11일 총 532부를 배부하여 401부(회수율 75.4%)가 회수되었으며, 이 중 부실 기재된 것을 제외한 393부를 통계자료로 이용하였다. 연구 결과 영양(교)사의 직무변인 평균 점수는 커뮤니티 3.38/5.00점, 자율성 3.32/5.00점, 가치 3.07/5.00점, 공정성 2.93/5.00점, 보상 2.91/5.00점, 업무량 2.21/5.00점 순으로 나타났다. 직무소진 중 정서적 소진은 3.50/5.00점, 냉소는 2.69/5.00점이었다. 직무열의 중 활력은 2.79/5.00점, 헌신은 3.22/5.00점, 몰입은 3.21/5.00점으로 나타났다. 직무변인, 직무소진 및 직무열의의 변인 간의 상관관계에서, 업무량은 정서적 소진(P<0.001)과 냉소(P<0.01)와는 유의한 음(-)의 상관관계를, 활기와는 유의한(P<0.01) 양(+)의 상관관계를 보였다. 자율성은 정서적 소진(P<0.001)과 냉소(P<0.01)와 유의한 음(-)의 상관관계를, 활기, 헌신, 몰입과는 유의한(P<0.01) 양(+)의 상관관계를 보였다. 보상은 정서적 소진(P<0.001)과 냉소(P<0.01)와는 유의한 음(-)의 상관관계를, 활기, 헌신, 몰입과는 유의한(P<0.01) 양(+)의 상관관계를 보였다. 커뮤니티는 정서적 소진(P<0.001)과 냉소(P<0.01)와는 유의한 음(-)의 상관관계를, 활기, 헌신, 몰입과는 유의한(P<0.01) 양(+)의 상관관계를 보였다. 공정성은 정서적 소진(P<0.01)과 냉소(P<0.01)와 유의한 음(-)의 상관관계를 보였으나, 활기, 헌신, 몰입에 대해서는 유의한(P<0.01) 양(+)의 상관관계를 보였다. 가치는 정서적 소진과 냉소와는 유의한(P<0.01) 음(-)의 상관관계를, 활기, 헌신, 몰입에 대해서는 유의한(P<0.01) 양(+)의 상관관계를 보였다. 직무변인이 직무소진, 직무관여에 미치는 영향력에서, 정서적 소진에는 업무량(${\beta}=-0.521$, P<0.001), 보상(${\beta}=-0.164$, P<0.001), 공정성(${\beta}=-0.129$, P<0.01), 가치(${\beta}=-0.094$, P<0.05)는 음(-)의 영향력을 주었고, 냉소에는 보상(${\beta}=-0.137$, P<0.05), 가치(${\beta}=-0.325$, P<0.001)가 유의하게 음(-)의 영향을 주었다. 활력에는 업무량(${\beta}=0.090$, P<0.05), 보상(${\beta}=0.165$, P<0.001), 가치(${\beta}=0.392$, P<0.001)가 양(+)의 영향을 주었고, 헌신에는 보상(${\beta}=0.147$, P<0.01), 가치(${\beta}=0.443$, P<0.001)가 양(+)의 영향을 주었으며, 몰입에는 가치(${\beta}=0.444$, P<0.001)가 양(+)의 영향을 주었다. 이상의 결과를 살펴볼 때 영양(교)사는 일과 보상이 적절하고 자기 일에 대해 가치를 느끼면 심리적 이탈감이 낮아지고 자발적으로 직무를 수행하는 열의가 높아지지만, 자신의 직무에 대해 공정하지 않은 처우를 받게 되면 정서적으로 지치고 스트레스를 받게 된다. 따라서 1일 2식을 제공하는 고등학교 영양(교)사의 과중한 업무량을 줄이는 방안을 마련하고 공정한 처우와 업무에 따른 적절한 보상이 이루어지도록 하며, 무엇보다 영양(교)사들이 자신의 직무 가치에 만족할 수 있도록 해야 하겠다.