1) CapeDwarf uses infinispan as the data grid. Does CapeDwarf use the cache store of infinispan to do permanent data store? Which cache store library it uses(LeverlDB?Cassandra?HBase?)?
By default it uses file store, but you can easily configure any other store -- see Infinispan / WildFly docs for this.
2) Using the clustered version CapeDwarf seems very easy from the users' document. But how does CapeDwarf do that?
This is default clustering behavior from WildFly clustering.
CapeDwarf simply reuses existing JGroups stack.
Which machine is the central control? How to manage that?
There is no central control -- you control, via configuration, how many replicas are there for any cache, etc.
The best way to get more details on this is to read the Infinispan or WildFly clustering docs,
as that's what CapeDwarf uses, there is no hidden magic on top of it.