Performance Optimization techniques HBase

[Neha]

I am a newbie in HBase. I am working on importing large scale data. 
What are the general optimization techniques/ways that should be used, while
developing such a code? 
We are using HBase Java API (Mapper/Reducer architecture). 
Also, we don't want to do the alterations at byte level, using HFiles (as done
by ImportTsv).
Could someone please suggest some techniques/ways to do this efficiently.