Analysts commonly use execution traces collected at runtime to understand the behavior of applications running on parallel and distributed systems. These traces are inspected post mortem using various visualization techniques that are generally incapable to scale properly for many events. This issue, mainly due to human perception limitations, is also the result of the screen size, which prevents the proper drawing of many graphical objects. Several visualization techniques tackle these issues by reducing the representaton complexity, using visual or data aggregation, or even clustering. Nevertheless, these solutions have drawbacks that hinder the analysis. We first evaluate existing trace visualization techniques using different criteria, involving how they are readable, their fidelity to represent the trace content without modifying its meaning, and so on. The objective is to determine which factors are responsible for the issues mentioned above. Second, we show how the combination of several aggregation techniques, data and visual, through a coherent and uniform treatment on spatial and temporal dimension, helps us to fulfill better the different criteria. This example enable us to claim the necessity of formalizing an aggregation methodology to provide decent spatiotemporal trace overviews for performance analysis.