Kassiano José Matteussi

Advisor Prof. Claudio Geyer

Interests and Activities

I am studying the resource consumption of Big Data applications on clouds. I am searching to avoid interference levels such as resource contention (CPU, Memory, Disk, Network, Energy) in Big Data clusters, caused by concurrently resource usage. At my first attempt to address a small slice of the problem, I tried to improve the resource allocation (disk at the first moment) to try to reduce application completion time using static allocations of resources. The obtained results sound good, and the next step is towards to do resource allocations for a set of Big Data applications dynamically. However, if necessary, I am strongly motivated to work in one of other research interest topics for work abroad (full doctoral, sandwich, double degree) experience.

Big Data

Prof. Cláudio Geyer

The explosive growth of data shaped the term called Big data which became a powerful research line that attracts notable attention from the most varied organization in academia and industry around the world. Following this trend, we employ several research efforts in data-intensive systems and processing, such as batch and stream processing applications over main MapReduce implementations like Flink, Spark and Storm. Thus, we have proposed strategies to use hybrid processing for big data analytics, techniques for resource and application scheduling, machine and deep learning solutions, system monitoring, MapReduce simulators, varied architectures and middlewares, performance prediction and analysis, the evaluation of ARM technologies and “green” solutions.

Students:

• Real Time Processing
  Julio C. S. dos Anjos
• Resource Management for Big Data Systems and Applications on Clouds
  Kassiano José Matteussi
• Big Data Architecture for Machine Learning Applications
  Breno Fanchiotti Zanchetta
• Monitoring of Computational Resources in Big Data Processing
  Jobe Diego Dylbas dos Santos

Cloud Computing

Prof. Alexandre Carissimi, Prof. Cláudio Geyer, and Prof. Philippe Navaux

Cloud Computing is considered as a significant advancement that delivers IT infrastructure, resource sharing, scalability, elasticity, and applications as services. These characteristics created an environment even more attractive to test data centers' capabilities (e.g., hardware, database, user-interface, and cloud-based applications) for the most general contexts, such as social, educational, development and Health areas.

Students:

• HPC Applications Running on Cloud Computing Environments
  Anderson M. Maliszewski, Eduardo Roloff, Emmanuell Carreno
• MicroClouds for Big Data Systems
  João Gross
• Cloud simulation over SimGrid
  Astélio José Weber Júnior

Computer Architecture

Prof. Philippe Navaux

Computer Architecture has been responsible for a scientific revolution. Its continual development improved computational power, increasing the range of problems that can be treated. The introduction of pipelines, increased frequency of operation, out-of-order execution, and branch prediction are an important part of the technologies introduced up to the end of the 20th century.

Students:

• Collaborative Execution and Mixed Precision
  Gabriel Freytag
• Scheduling and Memory Management
  Arthur Mittmann Krause, Matheus da Silva Serpa, Valéria Soldera Girelli
• Resilience
  Francis Birck Moreira

High Performance Computing

Prof. Lucas Schnorr and Prof. Philippe Navaux

High Performance Computing provides tools for solving complex, compute-intensive problems. As the field advances to the exascale, a processing power of 10¹⁸ floating point operations per second, many challenges arise for running applications efficiently and reliably.

Students:

• Parallel I/O for HPC
  Jean Luca Bez, Pablo José Pavan
• Application Performance Analysis and Optimization
  Vinícius Garcia Pinto
• Power and Energy Analysis
• Simulation of HPC Systems

FoG & Edge Computing

Prof. Cláudio Geyer and Prof. Philippe Navaux

Ubiquitous and pervasive computing or ubicomp also as called Internet of Things (IoT) describes basically that the devices (e,g., sensors, computers, tablets, cameras and so on) "things" are everywhere, interconnected and interacting in different ways. The underlying IoT technologies include many open challenges in the following areas such as Internet, middleware, operating system, mobile devices, sensors, microprocessors, new I/O and user interfaces, networks, mobile protocols, location and positioning and new materials [1]. [1] "Internet of Things Global Standards Initiative". Harvard Business, 2015.

Students:

• Quality of Context for the Internet of Things
  Hélio Carlos Brauner Filho
• Fog Computing/Edge Computing
  
• Performance Evaluation of Big Data Applications on FoG and Edge Environments
  Jorge Ximendes Silva Júnior
• Auto-Scheduling using Blockchain
  Rodrigo Barcelo dos Passos

Petrobras Exaflop

Paralelismo, Portabilidade e Eficiência de Aplicações Sísmicas de Larga Escala em Arquiteturas Exaflop

Oil and gas have been among the essential commodities for over a century. To improve their extraction, companies invest in new technology, which reduces extraction cost and allow new areas to be explored. Computing science has also been employed to support advances in oil and gas extraction technologies. Techniques such as computing simulation can be used to evaluate scenarios quicker and with a lower cost. Several mathematical models that simulate oil and gas extraction are based on wave propagation. To simulate with high performance, the software must be written considering the characteristics of the underlying hardware. Therefore, new architectures require changes to the software to support the new hardware features. The performance of the newest architectures depends not only on thread-level parallelism but also on parallelization based on vector instructions. Several challenges must be overcome by the software to use both types of parallelization. This research project explores several optimization techniques to improve the performance of a wave propagation model provided by Petrobras, a multinational corporation in the petroleum industry.

HCPA Retinopatia

Detecção de Retinopatia Diabética usando Aprendizado de Máquina em GPUs

Diabetes mellitus is a chronic disease associated with hyperglycemia due to failures in the secretion of insulin and the regulation of glucose production. It has an increasing prevalence, currently affecting 12% of the Brazilian population. The sugar concentration in the veins of the eye cause lesions designed diabetic retinopathies, which in turn will gradually cause blindness if left unattended. As a prevention method, retinographies are taken and analyzed by ophthalmologist doctors to evaluate whether a person suffers from this symptom, thus redirecting the patient to additional care or surgery. However, as the population grows, the demand for large scale health care systems becomes infeasible. Fully specialized doctors are expensive, and even resources for technicians to perform simple population screenings are limited. In this context, machine learning offers a prime opportunity to perform detection on a large scale within a limited budget. The retinography would be first analyzed by a convolutional neural network, which serves as a filter, thus reducing the number of ophthalmologists required.

Green-Cloud

Computação em Cloud com Computação Sustentável

• Project Information
  http://www.inf.ufrgs.br/greencloud/

Green-Cloud

Computação em Cloud com Computação Sustentável

• Project Information
  http://www.inf.ufrgs.br/greencloud/

ReDaS

Reproducible Data Science

• Reproducible Data Science
  https://www.inria.fr/en/associate-team/redas/
• ReDaS Associated Team
  https://associatedteam.gitlabpages.inria.fr/redas/