You might want to first read about various distributed file systems like andrew, CODA, intermezzo, before you start designing your own.
NFS is also a distributed file system. You have to ask yourself what are the goals of your distributed file system? Why NFS is not sufficient? (or is it?)
Amitay.
On Mon, 2003-03-24 at 00:43, Nikhil Karkera wrote:
Hi,
We need to create a distributed file system as part of an academic project. We'll be making the system an abstraction over the GNU/Linux FS. The user needs to start up a shell that we provide, and then he is on the DFS. The DFS is spread across about 8 machines in a LAN.
Since this is an abstraction over an existing FS, implementation details a not relevant. However, we were wondering over what would be a sound policy of data distribution across the machines? If a user creates a new file, how does the system decide on which machine to actually store it? Storing on the m/c that has maximum free space seems logical, but at the same time a lame policy (or is it the best way?).
Currently we are not considering fault tolerance, i.e. each file is stored at a single place (no duplicates). However, if anyone could, please do shed some light on that as well.
Warm Regards, Nikhil.
PS: We are using PVM and pthreads for programming. _______________________________________________________________________ Odomos - the only mosquito protection outside 4 walls - Click here to know more! http://r.rediff.com/r?http://clients.rediff.com/odomos/Odomos.htm&&o...