Minor feature support
---------------------
Hadoop
- namenode failover w/ shared dir mechanism
  - start w/ 2.1.1-beta

Ganglia
- if user wants aid in monitoring

Hbase
- multiple masters

Spark
- multiple masters
- "PER_NODE" variables, not correct

Storm + Trident

Tachyon
- fault tolerance (zookeeper + multiple masters, etc.)
- multiple layer hierarchy support
- localfilesystem underfs support

ideas
-----
failover support
- if too many nodes fail, recover through fake re-bringup?
- e.g. rank 10 node dies, have extra nodes to re-take over rank 10
- need to re-arch to make all nodes get rank based off file instead of actual rank
- would need daemon-ish thing to monitor situation, adjust config files, etc.?
