Current students

IT systems are becoming more and more complex and composed by a multitude of layers, each one with its own configuration comprising hundreds of parameters with interacting effects.
The traditional solution to this performance optimization problem consists in manually representing the IT system with stochastic models like queueing or Petri networks, which can then be used to simulate the behaviour of the system when exposed to a specific workload.
By varying some configuration parameters of the model it is possible to understand its behaviour. However, these approaches are limited in their accuracy and it is impossible to model the effects of many configuration parameters.
The scope of my research project is thus to use machine learning models to describe the performance behaviour of IT computer systems, together with an optimization algorithm to select the best configuration.