New and improved Hadoop port for Mesos

[Vinod Kone <vi...@gmail.com>]

We are excited to announce that we've recently committed a new and improved
Hadoop port for Mesos. The new port doesn't require patching Hadoop (i.e.,
it's a self contained Hadoop contrib) and let's you use existing Hadoop
schedulers (e.g., Fair Scheduler or Capacity Scheduler)! The tradeoff,
however, is that we don't take as much advantage of the fine-grained nature
of a Mesos task (i.e., there is no longer a 1-1 mapping between a Mesos
task and a map/reduce task*).

The new port has been committed to trunk at  r1439225.  For now, this patch
is still considered alpha, but we'd love to get some feedback!

To test the Hadoop port locally:

$ cd mesos
$ ./bootstrap
$ mkdir -p build && cd build # optional
$ ../configure
$ make # builds Mesos
$ cd hadoop
$ make hadoop-0.20.205.0 # alternatively, make hadoop-0.20.2-cdh3u3

To launch Hadoop on a real Mesos cluster, please run and follow the
instructions in "mesos/build/hadoop/TUTORIAL.sh".

Thanks,
@vinodkone


* We recognize that there are still opportunities to improve this, however,
so please let us know if you'd like to help out!