I've seen talks given by the IndexTank guys and their technology is pretty cool. ust to get an idea, they told me that Reddit's entire article submissions DB takes up around 4GB, and the comments DB, if it were indexed, would not be much larger.
Their business model seems to revolve around the fact that Lucene, Solr, SQL Server Full-Text search and other full-text indexing solutions just suck without serious customization, which can take months and cost thousands of dollars in consulting services, not to mention hardware costs. I believe they're handling all of Reddit's queries with one quarter of the hardware previously required (and unlike the previous version, it works).
