Search

Checkpoint/Restore techniques had been thoroughly used by the High Performance Computing (HPC) community in the context of failure recovery. Given the current trend in HPC to use containerization to obtain fast, customized, portable, flexible, and reproducible deployments of their workloads, as well as efficient and reliable sharing and management of HPC Cloud infrastructures, there is a need to integrate Checkpoint/Restore with containerization in such a way that the freeze time of the application is minimal and live migrations are practicable. Whereas current Checkpoint/Restore tools (such as CRIU) support several options to accomplish this, most of them are rarely exploited in HPC Clouds and, consequently, their potential impact on the performance is barely known. Therefore, this paper explores the use of CRIU’s advanced features to implement diskless, iterative (pre-copy and post-copy) migrations of containers with external network namespaces and established TCP connections, so that memory-intensive and connection-persistent HPC applications can live-migrate. Our extensive experiments to characterize the performance impact of those features demonstrate that properly-configured live migrations incur low application downtime and memory/disk usage and are indeed feasible in containerized HPC Clouds., This research was partially supported by the Spanish Government under contract PID2019-107255GB-C22, by the Generalitat de Catalunya, Spain under contract 2021-SGR-00478, and by the EU-HORIZON programme under grant agreement 101092646., Peer Reviewed, Postprint (published version)

The pandemic situation banned the presence of teachers and students in the classrooms disrupting the typical teaching process. Videoconferencing services have been used successfully for lectures, but they are insufficient for hands-on laboratories where the physical presence is a requirement (i.e. to access specific hardware). Enabling remote access to the equipment allows to overcome this issue, but the students must share resources (temporarily or physically) and lack full control on the equipment. This is not optimal for some subjects, such as the practical labs on datacenter management, which are increasingly being added into the IT higher education curriculum. In this paper, we present the design and implementation of a distributed laboratory environment, so-called Remote DasaLAB, to create and deploy virtual datacenters using the student's owned resources to support the practical training of students in remote (or hybrid) mode. This environment is flexible, cheap, easy to implement with available commodity resources (the students' computers) and provides full control over the platform. We also present a specific use case on leveraging Remote DasaLAB for the practical assignments of an Operating Systems subject, in which the students build a virtual datacenter composed of multiple nodes and can properly run a containerized application distributed in these nodes., This work was partially supported by Universitat Politècnica de Catalunya under grant G/22930/GEN: Projectes d’Innovació Docent 2021., Peer Reviewed, Postprint (author's final draft)

Background Therapeutic ceiling of care is the maximum level of care deemed appropiate to offer to a patient based on their clinical profile and therefore their potential to derive benefit, within the context of the availability of resources. To our knowledge, there are no models to predict ceiling of care decisions in COVID-19 patients or other acute illnesses. We aimed to develop and validate a clinical prediction model to predict ceiling of care decisions using information readily available at the point of hospital admission. Methods We studied a cohort of adult COVID-19 patients who were hospitalized in 5 centres of Catalonia between 2020 and 2021. All patients had microbiologically proven SARS-CoV-2 infection at the time of hospitalization. Their therapeutic ceiling of care was assessed at hospital admission. Comorbidities collected at hospital admission, age and sex were considered as potential factors for predicting ceiling of care. A logistic regression model was used to predict the ceiling of care. The final model was validated internally and externally using a cohort obtained from the Leeds Teaching Hospitals NHS Trust. The TRIPOD Checklist for Prediction Model Development and Validation from the EQUATOR Network has been followed to report the model. Results A total of 5813 patients were included in the development cohort, of whom 31.5% were assigned a ceiling of care at the point of hospital admission. A model including age, COVID-19 wave, chronic kidney disease, dementia, dyslipidaemia, heart failure, metastasis, peripheral vascular disease, chronic obstructive pulmonary disease, and stroke or transient ischaemic attack had excellent discrimination and calibration. Subgroup analysis by sex, age group, and relevant comorbidities showed excellent figures for calibration and discrimination. External validation on the Leeds Teaching Hospitals cohort also showed good performance. Conclusions Ceiling of care can be predicted with great accuracy from a patient’s, This study was partially funded by Secretaria d’Universitats i Recerca del Departament d’Empresa i Coneixement de la Generalitat de Catalunya (2020PANDE00148). The funding was used to collect data from the 4th wave in the MetroSud cohort. This work has also been supported by the Ministerio de Ciencia e Innovación (Spain) [PID2019-104830RB-I00/ DOI (AEI): https://doi.org/10.13039/501100011033, and by grant 2021 SGR 01421 (GRBIO) administrated by the Departament de Recerca i Universitats de la Generalitat de Catalunya (Spain)., Peer Reviewed, Postprint (published version)

Distributed applications present a different set of challenges from their centralized counterparts, such as consensus, clustering, or message propagation. Wetware is a distributed systems middleware that provides a unified solution to the most common challenges of building distributed systems, simplifying the development and deployment of distributed applications. Wetware can be used as a standalone program, setting up nodes as independent processes and accessing them via object capabilities; or it can be embedded into applications that make selective use of its functionalities. This thesis introduces a process executor for distributed applications as a novel component of Wetware, providing the means to run, manage, and communicate processes across a distributed cluster in a safe and performant manner through object capabilities. Process execution is based on WebAssembly, a binary instruction format that natively provides process isolation, portability, and performance; crucial characteristics of distributed processes in heterogeneous distributed systems. WebAssembly also introduces some challenges this thesis aims to solve, namely a lack of real parallelism and limited communication with elements out- side its sandbox. Furthermore, this thesis develops tools to manage said processes and provide the means for inter-process communication, and demonstrates their validity and effectiveness through the development of a real-world distributed application that runs on a Wetware cluster through the process executor. Next, the thesis analyzes the performance, resource usage, concurrency control mechanisms and scalability potential of the process executor for different work- loads. As a result, we demonstrate that the executor is capable of running multiple processes across nodes on a Wetware cluster, with work-bound performance akin to high-level programming languages and capable of managing IO-intensive workloads.

The operation of heat exchangers in a thermal system can be described by simple heat transfer equations. The heat exchange is normally defined by a global heat transfer coefficient, which takes into account internal and external convection and the characteristics of the pipe. This coefficient can be obtained analytically only for heat exchangers with standard and simple geometric characteristics otherwise leading to unacceptable inaccuracies when employed in lower-order models. An effective solution for calculating the global heat transfer coefficient for a particular heat exchanger consists of the full simulation of the heat exchange, including internal pipes and external fluid, through CFD&HT simulations. In this work, simulations of a coil heat exchanger are carried out, and the values obtained are compared with those of a reference work obtained experimentally [1]. In this work, the results obtained by means of different turbulence models present in the OpenFOAM software are compared., The author Jordi Vera has been financially supported by the Ministerio de Educaci´ on y Ciencia (MEC), Spain, (FPI grant PRE2018-084017). The author E.Schillaci acknowledges the financial support of the Programa Torres Quevedo (PTQ2018-010060)., Peer Reviewed, Postprint (published version)

Aquest treball de final de grau s'ha desenvolupat dins de l'empresa Barcelona Supercomputing Center per l'alumne Adrià Sabaté Creixell, el qual consisteix en la implementació d'un sistema de monitorització mitjançant una eina de software dels diferents serveis públics oferts pel BSC des d'un servidor contractat a un proveïdor de cloud. L'objectiu principal és l'obtenció d'una visió exterior general de l'estat dels diferents serveis., This bachelor thesis has been developed within Barcelona Supercomputing Center by the student Adrià Sabaté Creixell. The project consists of the implementation of a monitoring system, using a software tool, of the different public services offered by BSC from a server contracted to a cloud provider. The main objective is to obtain a general external vision of the state of the different services.

2023/2024

2023/2024

2023/2024

Tesi amb menció de Doctorat Internacional, (English) The convergence of High Performance Computing (HPC), Big Data (BD), and Machine Learning (ML) in the computing continuum is being pursued in earnest across the academic and industry. We envision virtualization and containerization technologies can be the basis for the convergence, because they reside as bridges between applications and infrastructures and provide well-known advantages. However, challenges remain for this convergence at the containerization level due to the diversity of applications and hardware heterogeneity: in the infrastructure layer, virtualization/containerization must feature complete isolation of the applications in a multi-tenant environment, seamlessly and efficiently provide different resources, allow agile and fine-grain dynamic resource provisioning to orchestrate resource sharing in those environments, also integrate HPC and Cloud scheduling and resource management techniques, while providing fault tolerance, energy efficiency, and scalability. Moreover, in the platform layer, virtualization/containerization must fulfill applications requirements of portability and reproducibility by allowing the definition of encapsulated and customized diverse software stacks, and the efficient creation/termination of those software environments on demand. The general research question of this thesis is: How to leverage virtualization/containerization in both the infrastructure layer and the platform layer to support efficiently the convergence of HPC, BD, and ML applications while taking advantage of heterogeneous HPC and Cloud resources? To answer that question, firstly, we enable deployments of HPC, BD, and ML applications using containers that allow the definition of encapsulated and customized software stacks for each application and provide seamless and efficient access to different resources through different configurations. This is the basis of the convergence. Secondly, the challenge is to understand the performance impact of the va, (Español) La convergencia de Computación de Alto Rendimiento (HPC), Grandes Datos (BD) y Aprendizaje Automático (ML) en el continuum informático se está persiguiendo con seriedad en el ámbito académico e industrial. Prevemos que las tecnologías de virtualización y contenerización pueden ser la base para la convergencia, porque residen como puentes entre las aplicaciones y las infraestructuras y proporcionan ventajas bien conocidas. Sin embargo, existen desafíos para esta convergencia a nivel de contenerización debido a la diversidad de aplicaciones y la heterogeneidad del hardware: en la capa de infraestructura, la virtualización/contenerización debe presentar un aislamiento completo de las aplicaciones en un entorno de múltiples inquilinos, proporcionar de manera fluida y eficiente recursos diferentes, permitir la asignación dinámica de recursos ágil y detallada para orquestar el uso compartido de recursos en esos entornos, también integrar HPC y Cloud planificación y técnicas de gestión de recursos, al mismo tiempo que proporciona tolerancia a fallos, eficiencia energética y escalabilidad. Además, en la capa de la plataforma, la virtualización/contenerización debe cumplir con los requisitos de portabilidad y reproducibilidad de las aplicaciones al permitir la definición de diversas pilas de software encapsuladas y personalizadas, y la creación/terminación eficiente de esos entornos de software bajo demanda. La pregunta de investigación general de esta tesis es: ¿Cómo aprovechar la virtualización/contenerización tanto en la capa de infraestructura como en la capa de plataforma para respaldar de manera eficiente la convergencia de aplicaciones de HPC, BD y ML mientras se aprovechan los recursos heterogéneos de HPC y Cloud? Para responder a esa pregunta, en primer lugar, habilitamos implementaciones de aplicaciones HPC, BD y ML utilizando contenedores que permiten la definición de pilas de software encapsuladas y personalizadas para cada aplicación y proporcionan acc, Postprint (published version)

Online machine learning (ML) inference services provide users with an interactive way to request for predictions in realtime. To meet the notable computational requirements of such services, they are increasingly being deployed in the Cloud. In this context, the efficient provisioning and optimization of ML inference services in the Cloud is critical to achieve the required performance and meet the dynamic queries by end-users. Existing provisioning solutions focus on framework parameter tuning and infrastructure resources scaling, without considering deployments based on containerization technologies. The latter promises reproducibility and portability features for ML inferences services. There is limited knowledge about the impact of distinct deployment schemes at the container-level on the performance of online ML inference services, particularly on how to exploit multi-container deployments and its relation with processor and memory affinity. In light of this, in this paper we investigate experimentally the containerization of ML inference services and analyze the performance of multi-container deployments that partition the threads belonging to an online learning application into multiple containers in each node. This paper shares the findings and lessons learned from conducting realistic client patterns on an image classification model across numerous deployment configurations, especially including the impact of container granularity and its potential to exploit processor and memory affinity. Our results indicate that fine-grained multi-container deployments and affinity are useful for improving performance (both throughput and latency). In particular, our experiments on single-node and four-node clusters show up to 69% and 87% performance improvement compared to the single-container deployment, respectively., This work was partially supported by Lenovo as part of Lenovo-BSC collaboration agreement, by the Spanish Government under contract PID2019-107255GB-C22, and by the Generalitat de Catalunya under contract 2021-SGR-00478 and under grant 2020 FI-B 00257., Peer Reviewed, Postprint (author's final draft)

El present treball exposa un estudi avançat de la resolució numèrica de les equacions multidimensionals de Navier-Stokes mitjançant mètodes computacionals. El projecte ha constat d’un detallat anàlisi teòric, a partir del qual s’han desenvolupat diferents simulacions de casos de referència amb el llenguatge de programació C++. Finalment, s’ha portat a terme una aplicació en el camp d’enginyeria de la climatització; en concret, en fluxos turbulents donats en la dinàmica de fluids de cortines d’aire. L’estudi de les equacions de Navier-Stokes ha sorgit de voler contribuir a la seva recerca científica degut a la seva complexitat. A partir de simulacions CFD, s’han obtingut solucions numèriques des de casos elementals a més avançats, podent-se endinsar en règims turbulents. Dels casos fonamentals, s’ha pogut verificar i validar la resolució numèrica considerant una varietat de geometries, condicions de contorn, dimensions de malla i esquemes numèrics. I dels casos més avançats, s’ha ampliat l’anàlisi a tres dimensions i a règims dinàmics turbulents, els quals no haguessin sigut possible sense la implementació de models de turbulència, ressaltant el model de Smagorinsky. A més, cal destacar d’ells el paper de la viscositat cinemàtica turbulenta en el tipus de simulació LES, que filtra els petits remolins de turbulència en l’espai i només resol directament les escales grans. L’aplicació desenvolupada simula un cas simplificat del flux turbulent que produeix la cortina d’aire, un equip que impulsa aire condicionat que té com a funció separar tèrmicament dos espais. En l’anàlisi realitzat, s’han avaluat la influència de la temperatura i la velocitat en els resultats, dels quals es conclou que una cortina d’aire isoterma és més energèticament eficient comparada amb una calefactada perquè presenta més pèrdues de calor, tot i que ajudi a mantenir millor la calor

Nuclear fusion, the process that could potentially bring clean, abundant, and virtually limitless power generation on Earth, has been a subject of intense research and exploration for decades. Among the scientific difficulties are breeding blankets, which play a crucial role in harnessing the energy generated during the fusion process. This study focuses on the analysis and exploration of magnetohydrodynamic (MHD) flows within the context of nuclear fusion breeding blankets. The thesis is structured into two sections: the first section is dedicated to the theory behind MHD, while the second section focuses on numerical simulations using developed MHD formulations. The numerical code employed in this thesis predominantly utilizes the phi-formulation, aligning with common practices in research applications. OpenFOAM and Foam-Extend are the primary tools used to simulate the MHD phenomena, with cases run for Hunt and Shercliff case setups. Furthermore, the cases are extended to higher Hartmann numbers, demonstrating excellent agreement with the analytical solutions for Hartmann numbers up to Ha=2500. Additionally, a small application to a U-bend case, representing scenarios commonly found in breeding blankets, is also presented.

2023/2024

2023/2024

El treball consisteix en el desenvolupament de codis en C++ per a la resolució de les equacions multidimensionals de Navier-Stokes, el que es coneix com a CFD (“Computational Fluid Dynamics”). A través de l’ús de mètodes numèrics, concretament de tècniques de volums finits i “solvers” d’equacions lineals, s’han obtingut els resultats per a una sèrie de casos d’estudi: Smith-Hutton, Lid Driven Cavity i Differentially Heated Cavity. Finalment, s’ha desenvolupat un cas propi que tracta sobre la propagació del fum originat en un incendi a l’interior d’un túnel. Al llarg del treball es desenvolupen les diferents equacions que s’utilitzaran en cada cas i es profunditza en cada un dels termes que les componen. Per altra banda, es mostra el procés de discretització temporal i espacial per a cada una de les equacions. En tots els casos estudiats, que disposen de solucions analítiques o de referència, es comparen els resultats obtinguts amb els de referència i es comprova que el codi desenvolupat té com a resultat uns valors molt pròxims als teòrics. El primer cas (Smith-Hutton) que es resol tracta sobre el transport d’una propietat en un fluid i, per resoldre’l, es fa ús de l’equació de la convecció-difusió. En aquest cas es parla tant de la resolució de l’equació com dels diferents esquemes convectius que es poden utilitzar. En el segon i el tercer cas es resolen les equacions de Navier-Stokes utilitzant el “Fractional Step Method”. En el segon cas (Lid Driven Cavity) s’analitza l’efecte que té una velocitat externa sobre un fluid que es troba contingut en un espai delimitat. En el tercer cas (Differentially Heated Cavity) s’analitza la convecció lliure a través de l’aproximació de Boussinesq i com aquesta modelitza els corrents que s’indueixen a causa dels canvis de temperatura en el fluid. En l‘últim cas s’introdueix el món de la turbulència i s’analitza com es propaga el fum en l’interior d’un recinte tancat. En aquest cas, s’analitza el model de Smagorinsky, un model de

A proper CFD simulation of air distribution within a room requires the correct representation of inlet conditions from ventilation grids. Those elements are usually characterized by blades with a certain inclination angle or even by a curvature, whose correct representation by finite volumes approaches requires the employment of very fine meshes. The coupling of those fine inlet domains with the coarser mesh of the environment can lead to unacceptable costs for common computational resources. In this work, an optimized methodology is presented to simulate the inlet flow conditions of an air diffuser within a bigger environment, like a room or a transport vehicle cabin. The method aims at reducing the computational cost of CFD simulations, by solving the velocity field of the diffuser in a local domain, and then mapping this velocity to the surrounding domain. Additionally, this allows us to solve the diffuser only once and reuse the velocity field multiple times for different room configurations., The author Jordi Vera has been financially supported by the Ministerio de Educación y Ciencia (MEC), Spain, (FPI grant PRE2018-084017). The author E.Schillaci acknowledges the financial support of the Programa Torres Quevedo (PTQ2018-010060)., Postprint (published version)