Skip to main content

COMPUTACION DE ALTAS PRESTACIONES EFICIENTE Y SEGURA PARA APLICACIONES DE SERVICIOS DE SALUD INTELIGENTES

PID2020-112496GB-I00

•
Funding agency name Agencia Estatal de Investigación
Funding agency acronym AEI
Program Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Call Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Call year 2020
State Plan 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Beneficiary centre UNIVERSIDAD AUTONOMA DE BARCELONA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 19
Encontrada(s) 1 página(s)

Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

Data-driven model for chronic kidney disease progression, a work in progress

  • Alvarez, Ana Candelaria|||0009-0001-0398-2797
  • Suppi, Remo|||0000-0002-0373-8292
  • Ibeas, Jose|||0000-0002-1292-7271
  • Balladini, Javier|||0000-0002-9769-7830
In medicine, data-driven models can be used to simulate disease progression and generate clinical decision support systems (CDSS). While artificial intelligence (AI) and machine learning (ML) models are common, their lack of traceability poses challenges in medical contexts, where transparency is crucial. This study aims to create a traceable data-driven model for Chronic Kidney Disease (CKD) progression without using AI or ML. The study will develop and validate a simulator based on this model with real-world data. This paper describes the development of the model, the processing of CKD patient data, the implementation of the simulator, and the validation of results.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

Efficiency analysis for AI applications in HPC systems. Cases study, K-Means

  • Rivas, Jose Maria
  • Wong, Álvaro|||0000-0002-8394-9478
  • Suppi, Remo|||0000-0002-0373-8292
  • Luque, Emilio|||0000-0002-2884-3232
  • Rexachs, Dolores|||0000-0001-5500-850X
Currently, many AI applications require large-scale computing and memory to solve problems. The combination of AI applications and HPC sometimes does not efficiently use the available resources. Furthermore, a lot of idle times is caused by communication, resulting in increased runtime. This paper describes a methodology for AI parallel applications that analyses the performance and efficiency of these applications running on HPC resources to make decisions and select the most appropriate system resources. We validate our proposal by analysing the efficiency of the K-Means application obtaining an efficiency of 99% on a target machine.
Dipòsit Digital de Documents de la UAB
Publication Artículo científico (JournalArticle)

A Metaheuristic Search Algorithm based on Sampling and Clustering

  • Harita Rascon, María de los Ángeles|||0000-0002-3732-4717
  • Wong, Álvaro|||0000-0002-8394-9478
  • Suppi, Remo|||0000-0002-0373-8292
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Luque, Emilio|||0000-0002-2884-3232
As optimization problems become more complicated and extensive, parameterization becomes complex, resulting in a difficult task requiring significant amounts of time and resources. In this paper we propose a heuristic search algorithm we call MCSA (Montecarlo-Clustering Search Algorithm), which is based on Montecarlo sampling, and a clustering strategy involving two techniques. Our objective is to apply MCSA, an inherently stochastic method, to address optimization problems. To assess its performance, we conducted an evaluation using classical benchmark optimization functions. Additionally, we leveraged the CEC2017 benchmark suite to comprehensively evaluate the algorithm, highlighting the pivotal role of the Exploration stage in our methodology. Subsequently, we extended our methodology to tackle a practical combinatorial problem, the Knapsack problem. This NP-Hard problem holds significant real-world applications in resource allocation, scheduling, planning, logistics, and more. Our contributions lie in parameterizing the Knapsack Problem to align with MCSA's parameters for reference indicators adjustment and achieving high-quality solutions, surpassing 90% in comparison to exhaustive methods such as branch and bound.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

An empirical method for processing I/O traces to analyze the performance of DL applications

  • Parraga Pinzon, Edixon Alexander|||0000-0002-5038-7827
  • León, Betzabeth|||0000-0003-1778-0237
  • Méndez, Sandra|||0000-0002-5793-1928
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Suppi, Remo|||0000-0002-0373-8292
  • Luque, Emilio|||0000-0002-2884-3232
The exponential growth of data handled by Deep Learning (DL) applications has led to an unprecedented demand for computational resources, necessitating their execution on High Performance Computing (HPC) systems. However, understanding and optimizing Input/Output (I/O) of the DL applications can be challenging due to the complexity and scale of DL workloads and the heterogeneous nature of I/O operations. This paper addresses this issue by proposing an I/O traces processing method that simplifies the generation of reports on global I/O patterns and performance to aid in I/O performance analysis. Our approach focuses on understanding the temporal and spatial distributions of I/O operations and related with the behavior at I/O system level. The proposed method enables us to synthesize and extract key information from the reports generated by tools such as Darshan tool and the seff command. These reports offer a detailed view of I/O performance, providing a set of metrics that deepen our understanding of the I/O behavior of DL applications.
Dipòsit Digital de Documents de la UAB
Publication Artículo científico (JournalArticle)

Methodology to Define a Static Allocation Mapping based on Memory Access Patterns and the Signature of MPI Applications in HPC Systems, Metodología para Definir una Política de Asignación Estática de Procesos basada en Patrones de Acceso a Memoria y la Firma de Aplicaciones MPI en Sistemas HPC

  • Enrique, Gerard|||0000-0002-4991-2196
  • Bruballa, Eva|||0000-0002-5094-3563
  • Suppi, Remo|||0000-0002-0373-8292
  • Wong, Álvaro|||0000-0002-8394-9478
  • Luque, Emilio|||0000-0002-2884-3232
  • Rexachs, Dolores|||0000-0001-5500-850X
Tools for identifying problems and improving MPI applications performance running on an HPC system require information from both the application and the system. In this work, we will focus on defining a methodology to analyze how memory usage affects an MPI application's performance running on an HPC system. This methodology will obtain valid and comparable data based on different memory access patterns, which will allow us to define key performance values used to characterize the HPC system behaviour facing these access patterns, as well as to characterize the Application Signature behaviour. This is obtained from Parallel Application Signatures for Performance Prediction (PAS2P) tool which obtains the representative phases of the MPI application, facing these same access patterns. With this methodology, we will be able to detect memory access application problems, suggest improvements and define a mapping policy for this application in this HPC system, in order to improve its performance and to determine limits to these improvements., Las herramientas para identificar problemas y mejorar el rendimiento de las aplicaciones MPI que se ejecutan en un sistema HPC requieren información de la aplicación y del sistema. En este trabajo nos centraremos en definir una metodología para analizar cómo el uso de la memoria afecta el rendimiento de una aplicación MPI que se ejecuta en este sistema HPC. Esta metodología obtendrá datos válidos y comparables basados en diferentes patrones de acceso a la memoria que permitirán definir valores clave de rendimiento utilizados para caracterizar el comportamiento de un sistema HPC frente a estos patrones de acceso y para caracterizar el comportamiento de la Firma de la Aplicación, (obtenida de la herramienta Parallel Application Signatures for Performance Prediction (PAS2P) que obtiene las fases representativas de la aplicación MPI) frente a estos mismos patrones de acceso. Con esta metodología, podremos detectar problemas de la aplicación de acceso a la memoria, sugerir mejoras y definir una política de mapeo para esta aplicación en este sistema HPC para mejorar su rendimiento y determinar los límites de estas mejoras.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

Modelling chronic kidney disease progression using ABM, a work in progress

  • Alvarez, Ana Candelaria|||0009-0001-0398-2797
  • Ibeas, Jose|||0000-0002-1292-7271
  • Balladini, Javier|||0000-0002-9769-7830
  • Suppi, Remo|||0000-0002-0373-8292
Chronic kidney disease (CKD) is a major health problem worldwide. In Spain, the incidence of CKD has increased nearly 17% in ten years, from 2010 to 2020; and in 2020 it had a prevalence of 20% in adults over 60 years. This disease has some characteristics that make it complex; for instance, CKD has a late diagnosis, because the symptoms appear when it is already advanced, and they are not directly related to kidney problems. Furthermore, the causes of the disease are multiple; the most common are arterial hypertension and diabetes. In this context, we believe that simulation and agent-based modelling can provide new tools for assisting professionals in the decision-making process. Therefore, we are proposing the design of a model for CKD progression in order to predict the appearance of renal failure and its evolution in patients through the different states of the disease. Our work is divided into four main stages; and in this article, we present the tasks that have been carried out so far in the first stage which are related to data analysis to define the complexity of the problem.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

ABM simulation focused on urban mobility

  • Gonzalez Cuevas, Antonio
  • Suppi, Remo|||0000-0002-0373-8292
The objectives of this research is to solve, through high-performance simulation, different mobility problems in cities with mixed traffic (public and private transport, emergencies, bicycles, scooters). To achieve these objectives, the research is oriented to develop new ABM models (Agent Based Models), integrated with Geographic Information Systems (GIS) in order to provide answers to different situations and mobility scenarios in big cities. We want to analyze different episodes and strategies (real or possible) to solve questions based on 'What if?' with variables such as mixed traffic, bicycles exclusive lanes, reduction of private vehicles lanes, among others. This work shows the first phase of ongoing research focused on the city of Barcelona, but the proposed methodology and models can be quickly extended to other cities. For this, we work with data captured in real traffic analysis processed with vehicle recognition algorithms based on AI (artificial intelligence) and the results are used as the basis for the ABM model based on GIS. The simulation is carried out by integrating initial real data with those generated by the simulation in order to analyze the mobility and their interactions.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

Patient-centric synthetic data generation, a new methodology for Chronic Kidney Disease

  • Alvarez, Ana Candelaria|||0009-0001-0398-2797
  • Ibeas, Jose|||0000-0002-1292-7271
  • Balladini, Javier|||0000-0002-9769-7830
  • Suppi, Remo|||0000-0002-0373-8292
Access to medical data is often restricted due to privacy and security policies. Synthetic data generation from real data is a widely adopted technique to address these limitations. This research presents a patient-centric, methodology for generating synthetic data, specifically designed for patients diagnosed with Chronic Kidney Disease (CKD). The key advantage of this proposal is its explainability and the traceability of the results, as it relies on statistics and data analysis rather than AI algorithms. The MIMIC-III clinical dataset serves as the foundation for generating synthetic patients in this study. This article details the data preprocessing and filtering applied to this dataset. Subsequently, synthetic data for CKD patients is generated using the proposed methodology. A comparison is then conducted between the synthetic data and the real data. Additionally, the synthetic data is compared with results obtained using the AI algorithm known as SMOTE. Generally, the metrics for the synthetic data generated by SMOTE are slightly superior. However, the results obtained with the proposed methodology exhibit minimal deviations from the MIMIC data across most CKD stages.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

Agent-based modeling of urban traffic scenarios for improved priority vehicle mobility

  • Gonzalez-Cuevas, Antonio
  • Wong, Álvaro|||0000-0002-8394-9478
  • Suppi, Remo|||0000-0002-0373-8292
This study is oriented to analyzing urban mobility challenges, with a focus on improving traffic flow conditions for priority vehicles in complex urban environments. Leveraging Agent-Based Models (ABM) in conjunction with Geographic Information Systems (GIS), the research aims to simulate various traffic scenarios to enhance the efficiency of urban mobility, with a particular emphasis on the movement of priority vehicles. The investigation explores the crucial role of priority vehicles, including but not limited to ambulances, within urban settings. It examines factors such as traffic congestion, infrastructure design, and the impact of traffic management strategies on the mobility of these essential services. A key component of this research involves conducting "What if?" analyses under different traffic conditions, such as mixed traffic flows, dedicated lanes, and emergency route prioritization. These simulations aim to evaluate the potential outcomes of various urban planning strategies on the efficiency of priority vehicle movement and overall traffic flow dynamics. This implementation serves as a foundational step towards understanding and simulating urban traffic scenarios. In the future, the project will integrate priority vehicles into the model and move to a High Performance Computing (HPC) environment. This article outlines the current state of the traffic model, focusing on the NetLogo implementation, and previews the upcoming integration of priority vehicles and the transition to HPC. While initially centered around Barcelona, the methodologies developed through this study are designed to be adaptable for application in other urban areas. The ultimate goal is to contribute to the development of effective traffic management strategies that prioritize the mobility of priority vehicles, thereby enhancing public safety and operational efficiency across densely populated urban environments.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

KP01 Solved by an n-Dimensional Sampling and Clustering Heuristic

  • Harita Rascon, María de los Ángeles|||0000-0002-3732-4717
  • Wong, Álvaro|||0000-0002-8394-9478
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Luque, Emilio|||0000-0002-2884-3232
In the field of optimization, NP-Hard problems play an important role concerning its real-world applications, such as resource allocation, scheduling, planning, logistics, etc. In this paper, we propose a heuristic search algorithm based on Montecarlo along with a clustering strategy that analyzes density and performs k-means partitions to solve the classic binary Knapsack Problem (KP01). Our heuristic method, which was designed to solve combinatorial optimization problems, has evolved and can adapt to other optimization problems, such as the KP01 that can be organized in an n-Dimensional search space. Regarding the methodology, we substantially reduced the search space while the areas of interest were located in the clustering stage, which brings us closer to the best solutions. After the experiments, we obtained a high-quality solution, which resulted in an average success rate of above 90%.
Dipòsit Digital de Documents de la UAB
Publication Artículo científico (JournalArticle)

Impact of the Population Pyramid on the Emergency Department, Impacto de la Piramide Poblacional en el Servicio de Urgencias Hospitalarias

  • Galeano Galeano, Ramona Elizabeth|||0009-0007-0272-1281
  • Wong, Álvaro|||0000-0002-8394-9478
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Suppi, Remo|||0000-0002-0373-8292
  • Bruballa, Eva|||0000-0002-5094-3563
  • Epelde, Francisco|||0000-0001-5547-4025
  • Luque, Emilio|||0000-0002-2884-3232
A country's population pyramid can affect the quality and demand for emergency departments (ED) in several ways. EDs may need specialized resources and equipment to meet the population's needs. In this research, we present an analysis of scenarios through agent-based modeling and simulation. We analyzed the population pyramids of Spain, Argentina, and Paraguay to provide insight into how demographic structure impacts the length of stay (LoS) and the need for medical and nursing staff. This can help policymakers and health managers better plan health resources and services in each country. We verified the evolution of the parameters, length of stay (LoS), and the occupation of doctors and nurses depending on different scenarios, such as the age of the patients and the number of patients arriving, and how it can lead to saturation of the ED. Through several scenarios analyzed using simulations of the age pyramids of Spain, Argentina, and Paraguay, we conclude that the age pyramid of patients treated in ED affects the demand for services, the need for specialized care regarding human and material resources, and waiting time. Implementing measures to manage demand and optimize available resources is essential to ensure adequate patient care., La pirámide poblacional de un país puede afectar la calidad y la demanda de los servicios de urgencias hospitalarias de varias maneras. En esta investigación, presentamos un análisis de escenarios mediante modelado y simulación basados en agentes. Analizamos las pirámides poblacionales de España, Argentina y Paraguay para proporcionar información sobre cómo la estructura demográfica impacta la duración de la estancia (LoS) y la necesidad de personal médico y de enfermería. Esto puede ayudar a los responsables y a los gestores sanitarios a planificar los recursos y servicios sanitarios. Verificamos la evolución de los parámetros, la duración de la estancia (LoS) y la ocupación de médicos y enfermeras en función de diferentes escenarios, como la edad de los pacientes y el número de pacientes, y cómo puede conducir a la saturación de los servicios de urgencias hospitalarias. A través de varios escenarios analizados mediante simulaciones de las pirámides de edad de España, Argentina y Paraguay, concluimos que la pirámide de edad de los pacientes en urgencias afecta a la demanda de servicios, la necesidad de atención especializada, y el tiempo de espera. Implementar medidas para gestionar la demanda y optimizar los recursos es esencial para garantizar una atención adecuada al paciente.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

Agent Based Simulation as an Efficient Way for HPC I/O System Tuning

  • Encinas, Diego|||0000-0002-6948-9786
  • Naiouf, Marcelo|||0000-0001-9127-3212
  • De Giusti, Armando|||0000-0002-6459-3592
  • Bond, Roman|||0000-0001-8021-8621
  • Méndez, Sandra Adriana|||0000-0003-3228-0134
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Luque, Emilio|||0000-0002-2884-3232
Generally, evaluating the performance offered by an HPC I/O system with different configurations and the same application allows selecting the best settings. This paper proposes to use agent-based modeling and simulation (ABMS) to evaluate the performance of the I/O software stack to allow researchers to choose the best possible configuration without testing the real system. Testing configurations in a simulated environment minimizes the risk of disrupting real systems. In particular, this paper analyzes the communication layer of an HPC I/O system, more specifically the communication layer of the parallel file system (PVFS2). ABMS has been selected because it enables a rapid change in the level of analysis for modeling and implementing a simulator. It can focus on both macro- and micro-levels (how the aggregate behavior of system agents is born and analyzing their individual behavior).
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

Modeling Parallel AI Applications for Performance Analysis on Cloud Environments

  • Albert, Miquel
  • Wong, Álvaro|||0000-0002-8394-9478
  • León, Betzabeth|||0000-0003-1778-0237
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Luque, Emilio|||0000-0002-2884-3232
In high-performance computing (HPC) environments, efficient execution of AI applications is critical for optimal performance and resource utilization. In this work, we extend the PAS2P methodology to AI applications through message passing on HPC Cloud systems, defining the AI Application Model to describe their performance behavior. This extension identifies phases within AI applications, enabling analysis to focus on these phases instead of the entire application. By concentrating on them, we can better evaluate AI application efficiency, providing insights into system performance and guiding future optimizations for large-scale AI tasks on HPC infrastructure.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

Preliminary Comparison of Different EDs Performance, Using Simulation

  • Galeano Galeano, Ramona Elizabeth|||0009-0007-0272-1281
  • Wong, Álvaro|||0000-0002-8394-9478
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Suppi, Remo|||0000-0002-0373-8292
  • Bruballa, Eva|||0000-0002-5094-3563
  • Epelde, Francisco|||0000-0001-5547-4025
A reference model is a model of something that contains a fundamental objective or idea of something and can be established as a reference for multiple purposes. We performed an analysis based on a reference model following the Spanish Ministry of Health's document of standards and recommendations to achieve better care, efficiency, and uniformity in emergency services. This standard describes the guidelines and resources needed in hospital emergency services for better patient care. We also analyzed the standards of emergency services in the following countries: the United States, the United Kingdom, Germany, Canada, Paraguay, Argentina, the United Arab Emirates, and Turkey. The objective of the research is to analyze the efficiency of the Spanish model following its reference standard compared to the standards of other countries, to explore how the Spanish emergency service would work using the parameters of the emergency service standards compared to the standards of different countries, specifically the KPI we analyzed is the Door to the Doctor (DtD), through simulation. It was concluded that in some countries, DtD times improve compared to the Spanish reference standard, and in some cases, they worsen.
Dipòsit Digital de Documents de la UAB
Publication Capítulo o parte de un libro (BookPart)

A Customizable Agent-Based Simulation Framework for Emergency Departments

  • Mesas Cervilla, Francisco|||0009-0007-9399-6298
  • Taboada González, Manuel
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Epelde, Francisco|||0000-0001-5547-4025
  • Wong, Álvaro|||0000-0002-8394-9478
  • Luque, Emilio|||0000-0002-2884-3232
Emergency Departments (EDs) face increasing complexity due to rising patient demand, resource constraints, and the need for efficient service coordination. Traditional simulation models, while useful, cannot be easily adapted to a different hospital environments, making it difficult to transfer and scale solutions. Based on previous work, with a simulator working in a hospital, this work describes a modular Agent-Based Modeling and Simulation (ABMS) approach for increasing flexibility adaptation and reuse in ED simulations. The proposed technique, which deconstructs existing models into individual components, will allow hospitals with different workflows and operational constraints to construct customized simulations. To validate this methodology, we develop a structured, modular framework using NetLogo and Python. The suggested metasystem enables adaptive simulation-based decision assistance for emergency departments, which improves resource allocation and operational planning.
Dipòsit Digital de Documents de la UAB
Publication Artículo científico (JournalArticle)

Deep learning data handling, exploring file formats and access strategies

  • Parraga Pinzon, Edixon Alexander|||0000-0002-5038-7827
  • León, Betzabeth|||0000-0003-1778-0237
  • Méndez, Sandra Adriana|||0000-0003-3228-0134
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Franco Puntes, Daniel|||0000-0003-0002-7046
  • Luque, Emilio|||0000-0002-2884-3232
Accessing large volumes of data presents a significant challenge when finding the best strategies to manage the data efficiently. Deep learning applications require the processing of massive amounts of data, which implies a considerable access Input/Output (I/O) load on computer systems. During training, interaction with the I/O system intensifies as files are continuously accessed to read data sets. This persistent access could overload the file system, which, in turn, adversely impacts application performance and efficient storage system utilization. Several factors influence the I/O of these applications, and one of the most relevant is the variety of file formats in which datasets can be stored. The choice of file format depends on the use case, as each format defines how information is stored. Some file formats have features that promote efficient access to datasets during the training phase, which can improve the performance of deep learning applications. Likewise, it is also important that the format adapts to the context, in this case, to an HPC system with a parallel file system. We will propose an image preprocessing method for cases where performance improves with parallel file access. This method will transform image data sets from their original JPEG format to the more efficient HDF5 format. Thus, our research will focus on the importance of understanding the mode of data access, spatial and temporal patterns, and the level of parallelism in file access to determine whether it is advisable to change the storage format.
Dipòsit Digital de Documents de la UAB
Publication Artículo científico (JournalArticle)

Application of a sampling and clustering-based heuristic search algorithm to find an efficient staff configuration in an emergency department

  • Harita Rascon, María de los Ángeles|||0000-0002-3732-4717
  • Wong, Álvaro|||0000-0002-8394-9478
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Luque, Emilio|||0000-0002-2884-3232
  • Bruballa, Eva|||0000-0002-5094-3563
  • Epelde, Francisco|||0000-0001-5547-4025
Emergency Departments (EDs) are among the most complex areas in healthcare, requiring immediate medical attention for acute and urgent conditions. Optimizing staff configurations to reduce patient Length of Stay (LoS) and improve operational efficiency poses a significant challenge due to the combinatorial and high-dimensional nature of the problem. To identify the most effective staff configuration, we propose a heuristic optimization strategy that is based on the Montecarlo Clustering Search Algorithm (MCSA), which efficiently explores the multidimensional solution space. MCSA leverages an agent-based simulation (ABM) model that evaluates each proposed staff configuration under realistic operational conditions, providing Key Performance Indicator (KPI) feedback values related to each proposed staff configuration. Through this strategy, we explore staff configurations capable of handling patient volumes with varying acuity levels in an ED to optimize the LoS KPI. Results demonstrate that our methodology is capable to find a solution as a staff configuration that reduces LoS compared to a baseline, offering a computationally efficient and practical tool for decision-makers. We identified solutions by exploring less than 1% of the total search space, demonstrating the efficiency of the proposed approach in addressing complex optimization problems. This approach supports informed planning in healthcare environments while maintaining system feasibility and scalability.
Dipòsit Digital de Documents de la UAB
Publication Artículo científico (JournalArticle)

A model of checkpoint behavior for applications that have I/O

  • León, Betzabeth|||0000-0003-1778-0237
  • Méndez, Sandra|||0000-0002-5793-1928
  • Franco Puntes, Daniel|||0000-0003-0002-7046
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Luque, Emilio|||0000-0002-2884-3232
Due to the increase and complexity of computer systems, reducing the overhead of fault tolerance techniques has become important in recent years. One technique in fault tolerance is checkpointing, which saves a snapshot with the information that has been computed up to a specific moment, suspending the execution of the application, consuming I/O resources and network bandwidth. Characterizing the files that are generated when performing the checkpoint of a parallel application is useful to determine the resources consumed and their impact on the I/O system. It is also important to characterize the application that performs checkpoints, and one of these characteristics is whether the application does I/O. In this paper, we present a model of checkpoint behavior for parallel applications that performs I/O; this depends on the application and on other factors such as the number of processes, the mapping of processes and the type of I/O used. These characteristics will also influence scalability, the resources consumed and their impact on the IO system. Our model describes the behavior of the checkpoint size based on the characteristics of the system and the type (or model) of I/O used, such as the number I/O aggregator processes, the buffering size utilized by the two-phase I/O optimization technique and components of collective file I/O operations. The BT benchmark and FLASH I/O are analyzed under different configurations of aggregator processes and buffer size to explain our approach. The model can be useful when selecting what type of checkpoint configuration is more appropriate according to the applications' characteristics and resources available. Thus, the user will be able to know how much storage space the checkpoint consumes and how much the application consumes, in order to establish policies that help improve the distribution of resources.
Dipòsit Digital de Documents de la UAB
Publication Artículo científico (JournalArticle)

Scalable performance analysis method for SPMD applications

  • Tirado, Felipe|||0000-0003-2001-658X
  • Wong, Álvaro|||0000-0002-8394-9478
  • Rexachs, Dolores|||0000-0001-5500-850X
  • Luque, Emilio|||0000-0002-2884-3232
The analysis of parallel scientific applications allows us to understand their computational and communication behavior. One way of obtaining performance information is through performance tools. One such tool is parallel application signatures for performance prediction (PAS2P), based on parallel application repeatability, focusing on performance analysis and prediction. The same resources that execute the parallel application are used to perform its analysis, creating a machine independent model of the application and identifying its common patterns. However, the analysis is costly in terms of execution time due to the high number of synchronization communications performed by PAS2P, degrading performance as the number of processes increases. To solve this problem, we propose a model that reduces data dependency between processes, reducing the number of communications performed by PAS2P in the analysis stage and taking advantage of the characteristics of single program, multiple sata applications. Our analysis proposal allows us to decrease the analysis time by 29 times when the application scales to 256 processes, while keeping error levels below 11% in the runtime prediction. It is important to mention that the analysis time is not considerably affected by increasing the number of application processes.