Dear sir: It is said In version 3.2 r/w performance improved by ~2x compared to the previous 3.1 release<http://hadoop.apache.org/zookeeper/docs/r3.1.1/zookeeperOver.html#Performance>. I have diff the src of the tow version, and find out the difference of FileTxnLog.java, which uses BufferedOutputStream. Could you tell me some detail about the performance improvement. Like algorithm, coding, data structure optimization. Thank you for your help.