If performance is important to you, Look at the quote from a previous thread:
"HDFS is a file system for distributed storage typically for distributed computing scenerio over hadoop. For office purpose you will require a SAN (Storage Area Network) - an architecture to attach remote computer storage devices to servers in such a way that, to the operating system, the devices appear as locally attached. Or you can even go for AmazonS3, if the data is really authentic. For opensource solution related to SAN, you can go with any of the linux server distributions (eg. RHEL, SuSE) or Solaris (ZFS + zones) or perhaps best plug-n-play solution (non-open-source) would be a Mac Server + XSan." --nitesh Besides, I wouldn't use HDFS for this purpose. Rasit