Cassandra input support now working ...

[burtonator]

I have Cassandra input support working ... I need to test it some more but the general functionality is there.

I have a branch called burton-cassandra-support that I have to merge into default first.

It's somewhat maintainable as I'm just using the stock Hadoop+Cassandra InputFormat and then decorate it to look like a peregrine JobInput ... 

Output should be easy after this point.

The general idea is that you write just have a normal job but instead of reading from a file you read from:

cassandra://localhost:9160/mykeyspace/graph

which is just a URI for building the config used in Cassandra.

There is more work to be done of course:

- It would be nice to actually have Peregrine unit tests startup Cassandra, import data into it, then have Peregrine map over it...

- I haven't run any benchmarks.

- We're taking the key/value maps for cassandra records and mapping them to a new interface since EVERY 'record' in cassandra is key/value based.  It wastes a bit of CPU to do this but I can't think of an elegant way to do this without breaking a LOT of abstractions.

- We don't do any routing based on host name.. so data is just randomly read.  This is the FALLBACK case for when you're reading data from non-local machines but we should do something more intelligent.  Directly mapping by hostname is the first step but understanding network topology is probably required eventually.