In recent years, ultraviolet (UV) based advanced oxidation processes have been extensively studied for degradation of refractory organic pollutants in water and wastewater, and selection of an appropriate UV light source is an important issue. In this study, bench-scale tests were conducted on a mini-fluidic photoreaction system (MFPS) to determine the degradation kinetics of methylene blue (MB) by UV/peroxydisulfate (UV/PDS) process equipped with a low-pressure UV (LPUV), vacuum UV (VUV)/LPUV, or medium-pressure UV (MPUV) mercury vapor lamp. Results indicate that MB degradation by UV/PDS with various light sources all followed the pseudo-first order kinetics, and the photon fluence-based rate constant (kp,λ′) had a descending order of: VUV/LPUV/PDS ≫ MPUV/PDS > LPUV/PDS. Moreover, it is noted that phosphate buffer (PB) notably inhibited MB degradation: the kp,LPUV′, kp,VUV/LPUV′ and kp,MPUV′ decreased by 35.0%, 44.9% and 37.5% with the PB concentration increasing from 0 to 20 mM, respectively. The maximal decrease in kp,VUV/LPUV′ was ascribed to a strong competition of PB for VUV photons. Thereafter, pilot-scale tests were conducted to evaluate the practical performance of UV/PDS in terms of the electrical energy consumption per order (EEO). It was found again that the VUV/LPUV lamp was the optimal light source in UV/PDS for organic pollutant degradation. This study helps optimize the UV/PDS process for its practical application to water and wastewater treatment.