Update on Hadoop

I have moved on to Hadoop 2 (Cloudera’s CDH4) — and I *definitely* suggest NOT manually maintaining hadoop nodes like I previously posted. It is much easier with Cloudera Manager. I also moved on to Hypertable, effectively replacing HBase for all our tables. It is slightly faster (8ms vs 20ms in django for puts) and the enterprise support is good (see hypertable.org UpTime subscription). The python modules work a lot better as well.