HDFS Snapshot

[#1]

Can someone explain me about HDFS Snapshot. I know it is point in time copies of file system, does it mean it keeps another copy of data? Lets say i have consumed 10 TB out of 100 TB and taken HDFS snapshot from / directory.

1. Does it keep another copy of 10TB under /.snapshot?

2. If snapshot is only metadata, how does NN constructs data when datablocks are deleted from datanode? or woudn't it delete datablocks at all? 

I'm confused can someone explain this?

[#2]

1. No, snapshot is just for the metadata operation.

2. Once you make particular directory snapshottable, the blocks belonging
the underlying files never be deleted.

[#3]

In addition to Dice's notes, please also read the design and efficiency overview at http://archive.cloudera.com/cdh5/cdh/5/hadoop/hadoop-project-dist/hadoop-hdfs/HdfsSnapshots.html#Ove... It will help gain a better understanding of the feature.