Study of the data acquisition network for the triggerless data acquisition of the LHCb experiment and new particle track reconstruction strategies for the LHCb upgrade

The LHCb experiment will receive a major upgrade by the end of February 2021. This upgrade will allow the recording of proton-proton collision data at $\sqrt{s} = 14\ \text{TeV}$ with an instantaneous luminosity of $2 \cdot 10^{33}\ \text{cm}^{-2}\text{s}^{-1}$, making possible measurements of unprecedented precision in the $b$ and $c$-quark flavour sectors. For taking advantage of the increased luminosity provided, the data acquisition system will receive a substantial upgrade. The upgraded system will be capable of processing the full collision rate of $30\ \text{MHz}$, without any low-level hardware preselection. This new design constraint poses a non-trivial technological challenge, both from a networking and computing point of view. A possible design of a $32\ \text{Tb/s}$ data acquisition network is presented, and low-level network simulations are used to validate the design. Those simulations use an accurate behavioural model developed and optimised for this specific purpose. It is mandatory to optimise the reconstruction algorithms using a computing and physics approach, to perform the online reconstruction of the full $30\ \text{MHz}$ $pp$ collisions rate. A new parametrisation of the charged particles' bending generated by the dipole of the LHCb experiment is presented. The accuracy of the model is tested against Monte Carlo data. This strategy can reduce by a factor four the size of the search windows needed in the SciFi sub-detector. The LookingForward algorithm in the Allen framework uses this model.