Why do seqdumper and clusterdumper poduce output in local disk?

When I run the two commands in hadoop mode , the output are all produced in
the disk. Why is the ouput in the hdfs in hadoop mode to perserve a
consistence?