Machine availability is a key indicator for the performance of the next generation of particle accelerators. Availability requirements need to be carefully considered during the design phase to achieve challenging objectives in different fields, as e.g. particle physics and material science. For existing and future High-Power facilities, such as ESS (European Spallation Source) and HL-LHC (High-Luminosity LHC), operation with unprecedented beam power requires highly dependable Machine Protection Systems (MPS) to avoid any damage-induced downtime. Due to the high complexity of accelerator systems, finding the optimal balance between equipment safety and accelerator availability is challenging. The MPS architecture, as well as the choice of electronic components, have a large influence on the achievable level of availability. In this thesis novel methods to address the availability of accelerators and their protection systems are presented. Examples of studies related to dependable MPS architectures are given in the thesis, both for Linear accelerators (Linac4, ESS) and circular particle colliders (LHC and HL-LHC). A study of suitable architectures for interlock systems of future availability-critical facilities is presented. Different methods have been applied to assess the anticipated levels of accelerator availability. The thesis presents the prediction of the performance (integrated luminosity for a particle collider) of LHC and future LHC up- grades, based on a Monte Carlo model that allows reproducing a realistic timeline of LHC operation. This model does not only account for the contribution of MPS, but extends to all systems relevant for LHC operation. Results are extrapolated to LHC run 2, run 3 and HL-LHC to derive individual system requirements, based on the target integrated luminosity.