I am considering Gluster for my application. In my application, I will have 8-10 servers with 10 GigE and 8 clients also with 10 GigE. At any time, each client will be doing a linear read of a 500-1000MB file. I know ahead of time which files will be read by which client, and they are disjoint (i.e., some files will be read by client 1, others by client 2, etc.) To avoid contention, I would like to have some control over which file lives on which server. But I would also like to be able to increase the number of servers in the future, and have their number not necessarily be a multiple of the number of clients, which is why something like Gluster is of interest. So, is the Gluster "elastic hash" algorithm pluggable? As in, would it be straightforward to implement my own? This is an internal application, so I do control the relevant file names... Thanks! - Pat
