Aha. I guess "Big part of them" is really big, so that simply using HBase counters directly is not fast enough.
The answer would be yes, HBaseHUT is meant for such situations as well. Though it is usually used with more complex calculations than just counters.
Currently with HBaseHUT there're no calculations (agg functions and such) out-of-the-box, so the merge code (processing updates) have to be written by client code. In you situation this means:
1) while processing input data, you write records (with the help of HutPut.adjustRowKey() as shown on wiki) which contain values like: column->inc_delta. I.e. by which number to increase values stored in specific column
2) while reading data or during periodic compactions you have to provide UpdateProcessor implementation, which for a given iterable list of records merges deltas and calculates the final value
Please let me know if you have questions for any of these parts.
Btw, such UpdateProcessor implementation looks quite generic for me, we could include it in HBaseHUT lib. So don't hesitate to create a pull request in github once you implement it ;)
Alex