Português English
Contato

Tese de Doutorado de Jhonny Marcos Acordi Mertz


Detalhes do Evento


Aluno: Jhonny Marcos Acordi Mertz
Orientadora: Profª. Drª. Ingrid Oliveira de Nunes

Título: Adaptive Filtering and Sampling in Runtime Software Monitoring
Linha de Pesquisa: Engenharia de Software

Data: 09/12/2021
Horário: 14h
Esta banca ocorrerá excepcionalmente de forma totalmente remota. Interessados em assistir a defesa poderão acessar a sala virtual através do link: https://mconf.ufrgs.br/webconf/prosoft

Banca Examinadora:
– Prof. Dr. Leandro Krug Wives (UFRGS)
– Prof. Dr. Valter Vieira de Camargo (UFSCar)
– Prof. Dr. Baldoino Fonseca dos Santos Neto (UFAL)

Presidente da Banca: Profª. Drª. Ingrid Oliveira de Nunes

Abstract: Understanding behavioral aspects of a software system is an essential enabler for many software engineering activities such as monitoring choke-points, debugging, and self-adaptation. Despite the usefulness of collecting system data, it may significantly impact the system execution by delaying response times and competing with system resources. Thus, runtime monitoring has limited practical use to online analysis or support real-time changes and adaptations on the program behavior. The typical approach to cope with this is to filter portions of the system to be monitored and to sample data. However, the majority of the existing solutions for filtering and sampling are limited to recording high-level events or based on predefined configurations, which unnecessarily limits the information available for analysis (i.e. relevance and representativeness of the collected set of traces). As systems often have varying workloads, some approaches dynamically change filtering and sampling configurations at runtime. Although these approaches are a step towards achieving a desired trade-off between the amount of collected information and the impact on the system performance, they focus on collecting data for a particular purpose or may capture a sample that may not correspond to the actual system behavior. In this thesis, we increase the practical feasibility of software runtime monitoring by reducing the monitoring overhead and ensuring the relevance and representativeness of the collected traces. Therefore, we propose a solution to address the challenges of filtering and sampling of execution traces. In order to filter relevant execution traces, we propose a domain-independent and low-impact framework, called Tigris, which abstracts the reasoning related to monitoring from the particularities of each problem addressed by filtering relevant execution traces according to the goal of monitoring. To tackle the challenges of collecting a representative sample, we propose an adaptive runtime monitoring process to dynamically adapt the sampling rate while monitoring software systems. It includes algorithms with statistical foundations to improve the representativeness of collected samples without compromising the system performance. We evaluated both approaches with empirical studies to assess different aspects of the proposed solutions. The results show that our techniques can reduce the overhead of monitoring by filtering and sampling traces, and ensure relevance and representativeness of collected traces.

Keywords: Execution traces, monitoring, sampling, performance, logging, adaptation, self-adaptation.