Microbial genomes used in biotechnology applications are now routinely sequenced in groups rather than individually, in order to more unambiguously identify specific variations that are linked to phenotype. These 'paraphyletic' sequencing strategies certainly result in growing volumes of sequence data, but these in turn are dominated by the n-ary relations between genomes obtained from systematic comparison, classification, and network inference. In the worse case, relations can grow geometrically while the genomes grow arithmetically. Comparative genomics is increasingly becoming a question of taming the complexity of these n-ary relations, and requires rethinking analyses in terms of new distributed computing paradigms. We will discuss a number of examples of large-scale comparative genomics in biotechnologically interesting hemiascomycete yeasts, and see how the MapReduce and NoSQL paradigms can be used to rethink representation, querying, and analysis of large groups of closely related genomes. I will further illustrate how reference-agnostic systematic comparisons can produce comprehensive views of the genomes as a group, and can drive comparative inference of metabolic models.