Performance tuning in underlying OS file system?

[jiang licht]

Any reference about settings of the underlying OS file system to optimize 
HADOOP performance? For example, what file system and block size to use for 
underlying Linux system? Thanks!

Best regards,

Michael