Commercial Hadoop support. Brave.
I know it just sounds easy to support "another" OSS project, but Canonical are going to find Hadoop different. Facebook, Yahoo! and others worry about any patch that could threaten their data, and to cut a release someone needs to stress test it on a real -not a virtual- datacentre with 1500+ servers. the current problems we are worrying about are disk balancing on 12HDD servers, better sharing of idle cluster time with other applications (through better reporting), etc etc -stuff you only find a problem on big clusters.
This is why the current Hadoop recommended config is always a specific Sun JVM that everyone is happy with, a specific RHEL/CentOS OS image that has also been tested by the big players at the petabyte scale.
If Canonical end up producing yet another Hadoop branch, that's just going to cause trouble. But without the ability to get their patches into the core -which they currently lack, not having any committers- that's exactly what's going to happen.
What is good here is that they could tweak ubuntu to work better with hadoop, such as
* allowing us to turn IPv6 off
* not defaulting to adding an entry to /etc/hosts that maps the local hostname to 127.0.0.1. This may make sense for laptops, but not machines in a datacentre.


