The adoption of effective methods for performance and dependability evaluation is crucial for any organization that wants to stay competitive in business. Performance evaluation can be done in several ways. In particular, it can be done through measurements on the real system, or using models. In many situations, model-based evaluation is the only alternative. This is the case, for instance, when the real system (or a prototype) is not yet available or when it would be too costly to monitor it. Model-based performance and dependability evaluation can be carried out using simulation models or analytical models (e.g., Markov Chains, Petri Nets, Reliability Block Diagrams). In our group, we develop and apply methods based on monitoring, simulation, and analytical models for system performance and dependability evaluation.

Optimization is the process of finding the best solution under given circumstances. Many problems in science, government and business can be conviniently modeled as an optimization problem. Algorithms to solve optimization problems can be broadly classified into exact and heuristic algorithms. Exact algorithms ensure optimality of their solutions. However, many real-world problems are large or complex enough to be solved in an exact manner within a reasonable amount of time. Heuristics are the main alternative to solve this class of problems. Heuristic algorithms try to find "good enough" solutions in a considerably shorter time than exact algorithms. OptLab is interested in developing heuristics (e.g., genetic algorithms, simulated annealing) and exact algorithms (e.g., ILP, dynamic programming) for complex optimization problems.

Many optimization algorithms are time-consuming, hence running these algorithms on a parallel architecture is an interesting option to speedup the optimization process. However, not all parts of an algorithm can be parallelized, and the parallelization complexity depends, among other things, on the algorithm itself, the parallel architecture, and the input data. Obtaining the best parallel implementation is a challenge on its own. There exists many parallel architectures, such as: multicore processors, field programmable gate array (FPGAs), graphics processing units (GPUs), and so on. Our research group is interested in developing parallel algorithms that are many times faster than their sequential counterparts.

The Institute of Computing is the unit from Federal University of Alagoas (UFAL) that is in charge of developing research and teaching activities in the fields of computer science and statistics.