why we need hadoop for hypertable
It's not required to use Hadoop. Hypertable only requires to be ran on top of distributed filesystem. So, you can choose between Hadoop DFS, KosmosFS, GlusterFS, Ceph, and maybe some others (like MooseFS or Lustre).
Common benefits are determined by application requirements and existent infrastructure. From a troubleshooting point of view, community (especially Hadoop's) size is also the great factor.
You can also look into cassandra (http://cassandra.apache.org/) another implementation somewhat similar to Big Table.