Why don't use existing DFS?

[howard chen]

such as Coda or GFS (RHEL) ,

i think their performance or features will be more mature?

I have run the random example to generate 10GB data, seems currently
HDFS is the bottomneck?


regards,
howa