Browsing by Author "Ilvonen, Sami"

Filter results by typing the first few letters
Now showing 1 - 5 of 5
  • Models and methods for computational electromagnetic dosimetry
    (2009) Ilvonen, Sami
     Doctoral dissertation (article-based)
    The interaction between electromagnetic fields and the human body is a very complicated issue. In most cases it is not possible to measure accurately the electrical response of the human body to external sources. Because of this computational methods are used as an aid when determining the safety levels for human exposure to electromagnetic fields. In this thesis the field distribution caused by various sources is determined in different parts of the human body using detailed and anatomically correct human body models and computational methods. Both the distribution of electric currents induced by low-frequency magnetic fields and the absorption of radio-frequency fields are studied. The accuracy and reliability of the models and methods used is verified by comparing the acquired results to known closed-form solutions and calibration measurements. The obtained results can be utilised in the reliability analysis of computational methods used in electromagnetic dosimetry. Furthermore, some of the results are needed, for example, in the safety guidelines of medical personnel working close to magnetic resonance imaging scanners.
  • Monitoring parallel file system usage in a high-performance computer cluster
    (2023-03-20) Tollander de Balsch, Jaan
     Perustieteiden korkeakoulu | Master's thesis
    Many high-performance computer clusters, rely on a system-wide, shared, parallel file system for large storage capacity and bandwidth. A shared file system is available across the entire system, making it user-friendly but prone to problems from heavy use. Such use can cause congestion and slow down or even halt the whole system, harming all users who use the parallel file system. In this thesis, we investigate whether monitoring file system usage in a production system at CSC can help identify the causes of slowdowns, such as specific users or jobs. The long-goal at CSC is to build an automatic, real-time monitoring and warning system that system administrators can use to make decisions on alleviating the slowdowns. Specifically, we monitor the usage of the Lustre parallel file system with Lustre Jobstats feature in the Puhti cluster, which is a petascale cluster with a diverse user base. We explain the necessary details of the Puhti cluster and our monitoring system to understand the Lustre file system usage data. During the thesis, we discovered issues in the data quality from Lustre Jobstats. The issues affected identifiers in the data, making some data unreliable and limiting our ability to build an automatic, real-time analysis. Nevertheless, we obtained a feasible data set for explorative data analysis. We demonstrate 24 hours of monitoring data by visually demonstrating file system usage patterns at low and high-level. Furthermore, we show that we can use file system usage data to identify causes of relative changes in I/O trends, particularly large relative increases. Finally, we explore ideas for future work on monitoring file system usage with reliable data from longer periods.
  • Rinnakkaistettu FDTD-menetelmä sähkömagneettisissa säteilytehtävissä
    (2000) Ilvonen, Sami
     Helsinki University of Technology | Master's thesis
    FDTD (Finite-Difference Time-Domain) on paljon käytetty sähkömagneettisten ongelmien ratkaisumenetelmä. Monet käytännön kannalta mielenkiintoiset tutkimuskohteet, kuten monimutkaiset säteily- ja sirontatehtävät, ovat kuitenkin olleet perinteisesti laskennallisesti liian suuria ratkaistaviksi tällä menetelmällä tavallisia työasemia käyttäen. Eräs tapa lähestyä suurikokoisia ongelmia on käyttää rinnakkaistettua FDTD-menetelmää sekä rinnakkaistietokoneita tai työasemaklustereita. FDTD-hilan säännöllisyyden ansioista käytettävä laskenta-alue on helppo jakaa yhtenäisiin ja samansuuruisiin alialueisiin, jotka voidaan osoittaa erikseen kullekin rinnakkaistietokoneen laskentayksikölle. Koska alialueet eivät ole toisistaan riippumattomia, täytyy alueiden reunoilla käyttää viestinvälitystä. Jotkin rakenteet, kuten absorboivat reunaehdot ja tasoaaltolähde, vaativat vielä ylimääräistä tiedonsiirtoa alialueiden välillä. Tässä työssä on toteutettu ohjelmisto, joka käyttää rinnakkaistettua FDTD-menetelmää sironta- ja säteilytehtävien ratkaisemiseen. Ohjelmien suorituskykyä on testattu käyttäen työasemaklusteria. Käytettyjen menetelmien oikeellisuutta on arvioitu tarkastelemalla tasoaallon sirontaa dielektrisestä pallosta, sekä vertaamalla saatuja tuloksia tunnettuihin arvoihin.
  • Real-Time GPU Usage Alert Service on Pre-Exascale HPC Clusters
    (2024-07-31) Jiang, Songlin
     Perustieteiden korkeakoulu | Master's thesis
    Improving observability in large-scale distributed computing clusters has always been a complex problem, particularly in High-Performance Computing (HPC). Despite the growing popularity of GPU-accelerated jobs, traditional workload managers in HPC systems, such as Slurm, lack the feature for collecting GPU usage history at job levels. In addition, with increasing workloads that rely on extensive GPU resources, especially AI training jobs, GPUs have become the most power-consuming hardware in the HPC system, so it's essential to reduce resource waste on these devices. To address these issues, in this master's thesis, we design and implement a real-time GPU usage alert service on top of the Slurm-based job monitoring system for supercomputer systems, i.e., Puhti, Mahti, and the pre-exascale supercomputer LUMI (the fastest supercomputer in Europe according to TOP500 by June 2024) at the CSC - IT Center for Science. We aim to have complete control over the data pipeline and tailor it to fit the characteristics of HPC systems so that it can be performant. As a result, we design our own GPU monitoring metrics collection infrastructure from the libraries provided by multiple GPU vendors and an in-memory real-time alert status checker service with the help of database triggers and LISTEN & NOTIFY. We also develop an alert algorithm to spot inefficient jobs with a bit of usage. In addition, we benchmarked the alert service with random data under extreme conditions designed for pre-exascale supercomputers, and the whole system was stable enough. Finally, we deployed the entire system in production for Puhti and Mahti, and it had been working well for months before we submitted the thesis. The outcome of this master's thesis empowers supercomputer administrators at CSC - IT Center for Science to learn about sub-optimal GPU resource utilization for specific jobs in real-time, finding out the cause for them, thus improving energy efficiency and significantly reducing resource waste in HPC clusters.
  • Software Engineering Methods for Cross-Team Projects with Multi-Project Teams
    (2024-03-11) Karlsson, Robin
     Perustieteiden korkeakoulu | Master's thesis
    Many software engineering methods exist for different types of projects and team structures. However, in organizations where cross-team projects are common, and each person in a team is involved in multiple projects, existing software engineering methods are difficult to apply. Projects with several teams, services, and technologies often also have a complex software architecture, which can be challenging to manage and coordinate. The goal of the thesis was to find software engineering frameworks or practices that can be used for projects at CSC - IT Center for Science Ltd. In the case project in this thesis, a new feature was implemented for the web interface for the LUMI supercomputer. Implementing the feature required changes in the services of three different teams with a wide variety of technologies used. Canonical Action Research was used to identify issues in the existing processes, test potential solutions in the case project, and evaluate the result. In the literature review, no existing software engineering frameworks that directly could be applied in the case organization were found. Instead, certain software engineering practices were selected from commonly used software engineering frameworks and previous research. The software engineering practices evaluated focused mainly on improving communication and coordination between the teams. The findings of the thesis, based on the case project, indicate that communication practices are important in software engineering. Good communication channels were beneficial for synchronizing activities and progress within the project. For communicating software architecture, diagrams made discussing technical details about the services and their interactions significantly easier. A coordinator role in the team was used to better help synchronize knowledge and activities between the teams.