To complement Wagstaff's good description of a keystone predator (it's worthwhile considering the etymology of the term), it should be noted that the scenario he describes is just one of the potential situations in which a keystone predator maintains diversity in an ecosystem. In fact, it's not the most usual situation.
The example of Pisaster, provided as a good example of the dynamic, is not in fact a case where the non-predatory species are competitively equivalent. In fact, the mussels are the competitive dominants, and in the absence of the starfish the ecosystem rapidly becomes a mussel monoculture. However, the starfish preys preferentially on the mussel, thus keeping it's density down to reasonable levels and permitting competitively non-dominant species to flourish.
This situation where the keystone predator feeds upon the competitively dominant prey species is, in fact, the most commonly observed configuration of this kind of relationship.