On 2014-04-30 21:59, Matthew Miller wrote:
One more thing. :) And this one might be a little bit more out there,
but
maybe worth considering. We've proposed making a cloud image tailored
for
big data. Maybe, instead, we should produce _Docker_ images tailored
in that
way, and suggest running those on our Fedora Atomic host.
I'm hesitant about doing a docker-only image for big data. The main
reason is that a lot of the big-data applications (cassandra, hadoop,
etc.) store large amounts of state on the node which seems less ideal to
handle in docker than just a plain instance running atop a hypervisor.