Hadoop and Project Voldemort

Great article in two parts (part 1, part 2) co-authored by Elias Torres (a colleague and friend) on using Hadoop and Project Voldemort for managing very large datasets.

It was interesting to read about the deployment issues they had to deal with as I dealt with some of the same issues with the indices for the search engine at Feedster (the update cycle there was 10 minutes), and the issue with key lookup as I had to engineer that for the search engine.