Hi,

Well, I've coded a couple of distributed thingies in the past (related with stress-testing, mostly) and I can tell you that distributed computing isn't all about dispersing small chunks of jobs to clients that do some CPU processing, you can also apply the same idea to storage. SETI@Home-type projects are not all that common because they need computational resources only and that doesn't happen very frequently, and when it does it may be the case that the computation cannot be practically broken to small distributable units of work. But everyone needs storage by the terabytes these days.

I started working on a pet project about grid storage once but work caught up with me and I dropped it - the idea was something along the lines of having clients surrender a part of their disk to the grid system and being able to manage the entirety of that like a single virtual file system. Features like fault tolerance and resilience (distributing the same data to more than one clients and defining/configuring a redundancy policy), parallel access (fetching data from more thah one client at the same time), client statistical dependency rating (which clients tend to dissapear frequently from the network and how often), decentralized access (fetching data from the nearest client and not via a server), automatic aging and cleanup (deletion of items that aren't being access at all or live in the grid for a predetermined period of time) and security can be incorporated. When working, the grid storage system could offer upload/download/search services to clients according to their security credentials...but such a thing would really be of use only when deployed to an environment with lots of desktops that have disk space to spare.

Cheers,
NTG