Local magnetic ordering in artificial spin ices is discussed from the point of view of how geometrical frustration controls dynamics and the approach to steady state. We discuss the possibility of using a particle picture based on vertex configurations to interpret time evolution of magnetic configurations. Analysis of possible vertex processes allows us to anticipate different behaviors for open and closed edges and the existence of different field regimes. Numerical simulations confirm these results and also demonstrate the importance of correlations and long range interactions in understanding particle population evolution. We also show that a mean field model of vertex dynamics gives important insights into finite size effects. Introduction.-Spin ices are geometrically frustrated magnetic systems, where interaction energies are minimized by local arrangements of spins resembling the ice rule for the ground state of solid water, i.e., two-in, twoout spin configurations [1,2]. Spin ices display several interesting features including zero point entropy [3], spin freezing, and hysteresis [4]. The role of long range interactions in three dimensional spin ices is also interesting, as the long range dipolar interactions between spins lead to a state that can be described using the short range ice rules [5].