I've worked at Google since 2005 on a variety of projects: web indexing, web serving, high-performance networking, and cloud. I've spent time optmizing existing systems (MapReduce and the websearch serving system, for example). I've chased down some crazy bugs (like a misplaced vtable that caused ::DeleteTable to be called instead of ::name()) and I found those SSTables someone lost by running MapReduce over 9TB of core files.

I also found some time to build large storage systems. You can read about one we published in OSDI: Percolator (slides). It was used to build a new indexing system (named "Caffeine" externally) that processes documents individually instead of in large batches. The image above was our official blog's attempt at explaining the system. Apparently in Caffeine you become the nucleus to the web's electron cloud.