Traditional filesystems are implemented as kernel modules (in Linux, at least) and together with userland tools, can be mounted and made available to end users. HDFS is what’s called a userspace filesystem. This is a fancy way of saying that the filesystem code runs outside the kernel as OS processes and by extension, is not registered with or exposed via the Linux VFS layer. While this is much simpler, more flexible, and arguably safer to implement, it means that you don't mount HDFS as you would ext3, for instance, and that it requires applications to be explicit...
Very important point
Share this highlighthttp://www.safaribooksonline.com/a/hadoop-operations/59109/