Scientific visualization is the process of visually extracting meaning from scientific data. This broad definition implies a continuum of approaches. At one extreme, visualization is simply a mode of presentation of previous results: a table or graph in a publication, an image or video in a slide show or on the web. At the other extreme, visualization is an interactive tool for experimentation: an environment that fosters further discovery by visually validating hypotheses and suggesting new ones. In this talk, we will explore the latter scenario within the context of scientific petascale data, and present novel approaches to visualization that scale with data size and maximize information content.