Tech Talk from the 2010 Reflections|Projections Conference (see http://www.acm.uiuc.edu/conference/2010/), posted courtesy of ACM@UIUC.
Project Horizon started as a hackday project at Palantir to explore what analysis is possible with Big Data. Building on Palantir's mission of enabling analysis at a level people understand rather than constrained by nitty-gritty data-oriented interfaces, Horizon enables interactive analysis that can be used by people that think that a sequel query is a question about The Wrath of Khan.
Rather than "write a complicated query, run it overnight," Horizon helps users iterate in seconds, asking new questions based on the answers they get at each step. Horizon uses a similar-but-different API to MapReduce and a highly-optimized memory datastore to deliver results in interactive time. We'll run through examples of the kinds of analyses this enables, and discuss the NoSQL datastore we built to make it all possible.
All Comments