Hadoop is built to treat all nodes the same but we have two classes of servers
in the system at this point... one for scene-by-scene processing and another
for stack operations (BU, LPVS, etc).
An investigation should take place to evaluate potentially replacing Hadoop
with a solution that allows us to account for this use case.
Original issue reported on code.google.com by davehil...@gmail.com on 2 Jan 2014 at 3:51
Original issue reported on code.google.com by
davehil...@gmail.com
on 2 Jan 2014 at 3:51