http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
An Analysis of Efficient GPU Resource Sharing for Concurrent HPC Application Executions
Theodora Adufu,하지원,김윤희 한국통신학회 2023 KNOM Review Vol.26 No.1
Graphics Processing Units (GPU) have become a significant part of today’s HPC and cloud infrastructures as they offer users exponential computing capacity to accelerate the execution of diverse applications at lower infrastructural costs. The traditional GPU however, allows for the execution of only one application at a time, giving rise to issues of resource under-utilization and subsequently higher deployment costs. Until now, efforts to enable concurrent executions on the GPU have been software-driven however with the introduction of the MIG feature by NVIDIA, GPUs can now be partitioned to provide isolated resources to applications. This paper investigates the behavior of HPC applications executed on different MIG instances. We also explore the implications of sharing GPU resources at the hardware level, on the performance of each application.
Theodora Zachariadou,Savvas Zannetos,Stella Elia Chira,Sofia Gregoriou,Andreas Pavlakis 한국산업안전보건공단 산업안전보건연구원 2018 Safety and health at work Vol.9 No.3
Background: Workplace bulling is a pervasive phenomenon with negative consequences for the health of victims and the productivity of organizations. The aim of this study was to measure the prevalence and forms of workplace bullying among employees working at the public health-care sector of Cyprus using the Greek version of Leymann Inventory of Psychological Terror (LIPT) instrument. Methods: A translation process was followed from the French to the Greek version of LIPT questionnaire. Testeretest reliability expressed by Pearson’s correlation coefficient was 0.98 indicating excellent reproducibility. Internal consistency reliability assessed by Cronbach a coefficient was 0.87 suggesting high reliability. LIPT instrument was distributed among 403 employees working at the primary healthcare setting and at the largest public hospital of Cyprus with response rate of 73.4%. Results: Women comprised the majority of participants (71.4%). Mean age was 43.3 years. Prevalence of workplace bulling according to Leymann’s definition was 5.9%. Most common forms of bullying were “Being continuously interrupted” (17.2%) and “continuously being given new work assignments” (13.5%). Women were significantly more often exposed to at least one mobbing behavior than men within the previous year (49% vs. 35.7%, p = 0.038), whereas nurses were significantly exposed to at least one mobbing behavior as compared to physicians (53.3% vs. 31.4%, p = 0.004). Conclusion: This was the first study that examined the prevalence of workplace bullying in the public health-care sector by elaborating the Greek version of LIPT instrument. Results showed that workplace bullying is a common and complex phenomenon among health-care organizations.
Antidiabetic Activity of Sargassum hystrix Extracts in Streptozotocin-Induced Diabetic Rats
Theodora Linggaryati Gotama,Amir Husni,Ustadi 한국식품영양과학회 2018 Preventive Nutrition and Food Science Vol.23 No.3
The objective of this research was to determine the potential effects of Sargassum hystrix extracts (SHE) on the glucose levels, lipid profile, and pancreas of streptozotocin (STZ)-induced diabetic rats. SHE at 200, 300, and 400 mg/kg was administered orally to STZ-induced diabetic rats once daily for 15 days. Glucose levels, lipid profile, and weight of rats were measured in the normal state and on the 15th day. The histology of the pancreas was observed on the 15th day. The results showed that the preprandial and postprandial glucose levels in the group treated with SHE at 300 mg/kg were significantly reduced compared with those of the diabetes group. Additionally, the levels of triglycerides and cholesterol in the 300 mg/kg SHE group were significantly different from those in the diabetes group. However, the levels of high-density lipoprotein cholesterol and low-density lipoprotein cholesterol across the treatment groups did not have significant differences. Necrosis was found in all STZ-induced rats. SHE at a dose of 300 mg/kg had the best capability to lower the levels of preprandial and postprandial glucose and to prevent necrosis in diabetic rats.
Dynamic Memory Allocation for Scientific Workflows in Containers
Theodora Adufu(아두푸 테오도라),Jieun Choi(최지은),Yoonhee Kim(김윤희) 한국정보과학회 2017 정보과학회논문지 Vol.44 No.5
대규모 HPC 과학 응용의 워크로드가 전체 실행시간 동안 다양하게 변화하는 자원 요구사항을 갖게 되면서 특정 시점에 갑자기 요구사항이 증가하는(bursty) 형태가 되고 있다. 그러나 이러한 응용 워크로드를 고려하지 않고, 최대 자원 요구사항만을 반영한 가상 자원의 오버-프로비저닝은 과학 응용의 성능을 보장하지만 다른 응용이 사용할 수 없는 유휴 자원을 늘리는 문제로 남아있다. 본 논문에서는 OS-level 가상화 환경에서 응용의 자원 사용 패턴에 대한 프로파일링 데이터를 기반으로 메모리 자원 재구성 기법을 제안한다. 이는 유휴 상태의 메모리 자원을 신속하게 풀어주어 새로운 응용이 자원을 사용하여 수행할 수 있도록 한다. 본 연구에서는 경량화된 OS-level 가상화 시스템의 하나인 Docker에서 과학 워크플로우 응용을 이용하여 제안하는 알고리즘을 검증하였다. 실험을 통해 과학 응용을 실행하는 동안 컨테이너에 대한 메모리 할당 미세 조정이 전반적인 메모리 자원 활용을 향상시킬 수 있음을 보였다. 또한 응용의 메모리 사용 프로파일 데이터를 기반으로 하는 시뮬레이션 실험을 통해, 제안하는 동적 메모리 할당 기법을 사용하는 경우 대기 작업에 유휴상태의 메모리를 할당하여 전체 대기 작업의 수를 줄이고 시스템 작업 대기 시간이 줄어들었음을 보였다. The workloads of large high-performance computing (HPC) scientific applications are steadily becoming “bursty” due to variable resource demands throughout their execution life-cycles. However, the over-provisioning of virtual resources for optimal performance during execution remains a key challenge in the scheduling of scientific HPC applications. While over-provisioning of virtual resources guarantees peak performance of scientific application in virtualized environments, it results in increased amounts of idle resources that are unavailable for use by other applications. Herein, we proposed a memory resource reconfiguration approach that allows the quick release of idle memory resources for new applications in OS-level virtualized systems, based on the applications resource-usage pattern profile data. We deployed a scientific workflow application in Docker, a light-weight OS-level virtualized system. In the proposed approach, memory allocation is fine-tuned to containers at each stage of the workflows execution life-cycle. Thus, overall memory resource utilization is improved.
A Performance Benchmark of Cached Data Access Patterns on GPUs
Theodora Adufu,김윤희 한국통신학회 2022 KNOM Review Vol.25 No.2
GPUs have successfully accelerated general-purpose applications in recent times as they provide much higher instruction throughput and memory bandwidth to applications from different domains. There remains however a bottleneck in the performance of GPUs due to large memory latencies. Caches reduce off-chip memory traffic however, managing caches is difficult. With the new residency control feature introduced in the Nvidia Ampere architectures, users can now control how much data is resident in the cache. During co-executions of multiple applications, the caveat however is to identify which application requires data persistence and by how much. In this paper, we characterize workloads by throughput and data access frequencies and experiment with three co-scheduling scenarios to determine persistent cache allocations for optimum performance. We observed that when co-executing applications with different data accesses, L2 set-aside allocations should not be biased towards applications with persistent data accesses.