You can clone with
No one assigned
In order to make use of data localization pending updates processing can be performed on per-region basis with the help of Coprocessors API.
This (hopefully) would be faster than MapReduce implementation since no overhead for MR job infrastructure. Also we shouldn't make MR cluster a requirement for HBaseHUT.
Looks like will not be available soon for 0.20.XX-0.90.X versions, since Coprocessors just got going (on trunk only)