Data-flow performance optimisation of the ATLAS data acquisition system

Colliding particles at higher and higher energies has proven to be a fruitful avenue to expand our knowledge of nature. Results from high-energy physics experiments have led to the formulation of the Standard Model, which has been strikingly successful in describing the currently known fundamental particles and the interactions between them. Nevertheless, the Standard Model is necessarily an incomplete theory as it does neither account for gravity, nor provide an explanation to cosmological problems like the apparent existence of dark matter and the observed matter-antimatter asymmetry. New phenomena, not contained in the Standard Model, could be discovered by pushing the energy boundary further. Current high-energy physics experiments aim to observe these new phenomena and explore the electroweak symmetry breaking mechanism predicted by the Standard Model. These will necessarily be concealed within a huge background of already well known processes. Therefore, not only the energy, but also the collision rate needs to be pushed further to obtain statistically significant results. As a consequence, despite the fast-paced evolution of information technologies, experiments generally cannot afford to cope with the large data volume generated in the detectors. A clever trigger and data acquisition system has to be deployed in order to select the interesting interactions for permanent storage and discard the background. This is a challenging task: current experiments require systems that perform very narrow selections (accepting one in \mathcal{O}(10^{5}) events) almost in real-time. The performance of a trigger and data acquisition system is crucial in realising the full discovery potential of an experiment. Contrary to off-line data analyses that can be improved upon and executed multiple times until the best results are obtained, a trigger discarding potentially interesting data, will lose data permanently. Therefore, special care needs to be taken in designing and operating a trigger and data acquisition system, with the aim of maximising the amount of interesting interactions that can be recorded without introducing uncontrolled bias. This thesis focuses on the trigger and data acquisition system of the ATLAS experiment at the Large Hadron Collider. The challenging experimental environment is described in chapter 1, while the design and implementation of the trigger and data acquisition system are illustrated in chapter 2. Its key performance parameters and their impact on the discovery potential are studied in chapter 3. In particular the specific area of heavy flavours physics is considered as a case study. In chapter 4 the last stage in the data acquisition system, the data logger, is discussed and a new design that enables significant performance gains is presented.