if I say that in hadoop cluster system... A client has 10GB data to send to name node. After analysing the "packet capturing" actually the name node doesn't divide the data for data node, actually all the communication happens only between client and data node , after taking the IPs of data node from name node, client directly connect to data node and after that he decide , in how many partitions he is sending the data by dividing the blocks. And then the data is receives by the data node and during this time name node contact to the data nodes to be prepare of replicating these data.

So my question is that the by this process I realized that the data sent by client is receives by data node in serial manner and date node copies that data regularly and hand to hand. Then if I sending a file of 100GB and usually it takes 100 minutes but after using tha Hadoop my speed and time is consuming more than before because the data node receives the data serially from client . "IS THIS A MAJOR DISADVANTAGE OF HADOOP??"



Browse other questions tagged or ask your own question.