Hadoop Writing Process

client to request suitable name nodes and set the pipe line

write across the pipe line from first suitable data note to the last

get feed back from the pipe line, which is reported from the last to the first suitable data note

update meta data in name node

blocks are writing parallelly