Our Apekshit Sharma wrote up a nice blog on the work he did in HBASE-11927 making it so HBase does native instead of java checksumming when Hadoop Native Library is available saving mountains of CPU. Check it out https://blogs.apache.org/hbase/
St.Ack
