HBase and Haystack; all data LZO-compressed; very interesting approach to testing -- they 'shadow the real production workload into the test cluster to test before going into production'. This catches a 'high percentage' of issues before production. nice

Fascinating. Evading the Java GC by reimplementing a slab allocator, basically

Scribe logs events, "ptail" (parallel tail presumably) tails logs from Scribe stores, Puma batch-aggregates, writes to HBase.  Java and Thrift on the backend, PHP in front