HDFS name node has single point of failure in Hadoop 1(but not Hadoop 2)

Small or bigger files does not matter, as it divides the files into blocks (64 MB by default). But recommended to have 128 MB or 256 MB in production)

Secondary node has a different purpose. It cannot take over name node.

These are my answers with my opinion, please update as needed.
Does HDFS has a single point of failure? - in Hadoop 2 yes
Does HDFS has rack awareness?- Not sure about this
Is HDFS designed for lots of small files or bigger files? - It is designed for bigger files, not for small files
Can Secondary NameNode takeover as NameNode if NameNode fails? Yes.

hi Balaji,

All of your questions should have been already touched upon by the instructor, however here are short answers

Does HDFS has a single point of failure - Yes NameNode is the single point of failure
Does HDFS has rack awareness - Yes, the Hadoop admin will setup the rack awareness.
Is HDFS designed for lots of small files or bigger files - Hadoop doesn't work well with smaller files as it tends to increase the metadata info in fsimage which will occupy the RAM of NameNode.
Can Secondary NameNode takeover as NameNode if NameNode fails - The takeover can only be setup in Hadoop 2.0 and again the failover is not automatic.

Thanks

Hi Balaji,
Srinivasan has already these correctly.

HDFS in Hadoop1 has a single point of failure in the NameNode. (Hadoop 2 solves for this - NameNode is not a single point of failure in Hadoop 2).
HDFS is much better suited for large files. For small files the overheads of meta info and replication of relatively small files does not add much value.




thanks,
Sarath.