Facebook is now using Hadoop for realtime workloads, which is interesting in and of itself. (Cassandra was Facebook’s go-to for NoSQL data.) They’ve also extended Hadoop to make it a more effective realtime system. This white paper and slides detail why they chose Hadoop, what changes they made and what is left to be done.

Spynner looks like it could be a somewhat suitable replacement for Selenium, without the dependency on a server process. Supports JS, runs on Webkit.