Al wrote: > damn > in the entire collage there has to 4 or 5 terabytes going begging > > i find it hard to believe unix doesn't have a simple cmd line that says > "make all free network hard drive space one contiguous volume" ;) > > maybe mac os x server has some tricks up it sleeves??
Not likely. Such a system would be really hard to manage. What happens in your theoretical distributed file storage when a computer is shut off? A drive goes south? A user is doing a really disk-intensive task? The user simply deletes the files to make room for *their stuff*. Worse you and the user each write a 10 gb file to that disk that only has 10 gb left? There are distributed database systems that work quite well; with the appropriate setup I can write an Oracle application that will access data on systems literally around the world. (I've known people who do this, in fact) I could, with lots of programmatical Junkyard Wars-like bodgering, bash a distributed Oracle installation into what appeared like a large single drive, but the client software, Oracle Server, is a wee bit on the heavy side. (Ours takes up slightly over 1 gig of disk space, which is just the server and associated software, not the databases) and is a *wee* bit complicated to run. P2P filesystems like Kazaa and Gnutella manage to ditribute files around to different systems like what you're talking about but the latency in finding a specific file is horrible, and they scale even worse. Napster style models, with distributed filestorage and centralized file indexing work far better and scales far better than Gnutella, but a centralized index gives you a centralized point of failure. And none of these models deal well with systems that get turned off, and none of these will split a file too large to go on one node across more than one. (oh, I *could* do it on a sufficiently advanced Oracle setup, but that's a corrolary of Clarkes Law: Any sufficiently advanced OODBMS is indistinguishable from Magic. And if I could do that, Larry himself would be offering me the use of his Mig for the weekend and such ;-) Push come to shove, if I had to make one of these beasts, I'd make it using the Napster model with a honking fast database server for indexing, like Oracle or postgres running on a box doing nothing but that, with a real fast network pipes, along with some small client running on the 'network storage devices', aka everyone else's computer. That said, when alls done, it's easier and cheaper just to buy bigger drives or standalone NSD's for your network, which is why people use file servers instead of some distributed file system. When you have a million node filesystem, you have a million security holes into your server. Gives me heebie-jeebies thinking about it. and backups, YIKES! I don't *want* to think about backups! When sysadmins are very very bad they go to places like that when they die, and get to be sysadmins in hell, and get stuck with Windows. Windows 1.0. On a Packard-Bell. A broken Packard-Bell. -- Bruce Johnson University of Arizona College of Pharmacy Information Technology Group Institutions do not have opinions, merely customs -- G-Books is sponsored by <http://lowendmac.com/> and... Small Dog Electronics http://www.smalldog.com | Refurbished Drives | -- Check our web site for refurbished PowerBooks | & CDRWs on Sale! | Support Low End Mac <http://lowendmac.com/lists/support.html> G-Books list info: <http://lowendmac.com/lists/g-books.html> --> AOL users, remove "mailto:"; Send list messages to: <mailto:[EMAIL PROTECTED]> To unsubscribe, email: <mailto:[EMAIL PROTECTED]> For digest mode, email: <mailto:[EMAIL PROTECTED]> Subscription questions: <mailto:[EMAIL PROTECTED]> Archive: <http://www.mail-archive.com/g-books%40mail.maclaunch.com/> Using a Mac? Free email & more at Applelinks! http://www.applelinks.com
