Computer Engineering Thesis Presentation After a long time of hard work and dedication, a thesis in Computer Engineering was successfully presented aboutautomatic software systems’ failure recovery through the modeling and simulation of controllers in real time, held in the framework of Epidata’s academic cooperation activities.
The thesis was presented by Guillermina Galache, who presented it as part of a university-industry collaboration between the Group of Discrete Events Simulation, Department of Computer Science at the Faculty of Exact and Natural Sciences and Epidata’s Research, Development and Innovation Department, Argentinian company specialized in software architecture.
The work was also presented to the academic community through a scientific publication in JAIIO 2012, the prestigious Argentine Conferences on Informatic and Operative Research which are annually organized in the Software Engineering Symposium (ASSE).
The original results on different topics are presented and discussed with the assistance of argentinian and foreign professionals. The thesis of Ing Guillermina Galache consists of the submission of a software platform to enable automatic and autonomous systems adaptation to a degradation in its quality of service and was directed by Dr. Ing Rodrigo Castro (FCEyN/UBA, ETH Zurich) and Ing Osvaldo Clúa (FI/UBA).
This development is expected to contribute generating synergy between the disciplines of Software Engineering and Control Engineering in the context of quality assurance of software systems service, aiming to produce software systems more robust and reliable.
Since its creation in 2007, the Epidata’s Research, Development and Innovation area (R + D + i) has conducted projects together with local and international universities. The activities coordinated by the area include: Management Thesis, Research Cooperation with Universities, Academic Cooperation Agreements, Free Assistance Training at Universities and Innovation Observatory, among others.
Thesis Development Summary
Self-Healing methodology offers particular advantages when closed loop control techniques are applied.
However, one factor limiting their adoption in industrial environments is the strong dependence of each particular control strategy with the specific tool that implements it, hindering the convergence to a standard platform. To counter this situation, we have developed a generic tool to design and implement different control strategies in a flexible and modular context. Its three core functions: monitoring, control and performance are achieved by independent components integrated through a data network.
The controllers are implemented in a modeling and simulation environment based on the DEVS formalism (Discrete Event System Specification) to specify in a mathematically robust way controllers of any kind: continuous, discrete-time, discrete event and / or combinations thereof (hybrids). This facilitates the development of control strategies according to the System-In-The-Loop approach (system “in the loop”): the models are developed, tested and verified by actual simulations decoupled of the real controlled system (system “out of the loop”), then when the performance is acceptable, the inputs and outputs of the controller are simply “redirected” to the real system (system “in the loop”) for validation.
Through an iterative process, the simulation models become executed in real time making them directly on the final product. The effectiveness of the platform was demonstrated through the design of a simple on-off control type to adjust the CPU and memory consumption of a Java EE web application. The essays presented unexpected dynamics of the controlled application, and platform flexibility to tailor showed control incorporating new knowledge.
