choose which Hector's serializer for Cassandra performance?

[박용욱]

I have questions below.

1. In cassandra server, row key, column name and column value are saved in
byte[], aren't they?

2. If I call hector's mutator.addInsertion(rowKey, cfName, HColumn<String,
Integer>), does the transformation from String/Integer to byte[] occur at
client? or server?

3. If 2 occurs at server, then which Type should I use with HColumn<Type,
Type> to get the most performance of cassandra? client performance is out of
concerns.


4. with RF 3 and random partitioner, does read iterations(paging) with range
query work well? I mean whether row-keys order changes over iterations or
not? (CL is QUORUM)

Sencerely.
Colin.