128-bit storage, and boiling oceans
I was going through some old bookmarks the other night, and happened across an old gem on Jeff Bonwick's blog, "128-bit storage: Are You High?".
It's a great little bit of the web, and explains why Sun decided to jump to 128-bit storage for their new file system instead of 64-bit, along with a small detour about the theoretical limits of storage capacity and just how much it would take to fully-populate a 128-bit data pool.
Chances are your head will be spinning by some of the examples, but much is elaborated on in the comments, which shouldn't be missed. It really shows off the value employee and developer blogs can add for why some things are the way they are. I know it certainly opened my mind on the subject.
Posted by drunkenbatmanJanuary 26, 2005, at 05:22 PM
Comments (2)
Posted by: drunkenbatman at January 26, 2005 09:25 PM
Slightly vague, but think of it this way: imagine hard drive as little individual puddles of data. If you collect them all and pour them together so they can be accessed as one entity, you have a pool.
Let's say you need to work with a file that is 5 terabytes. They don't make a 5-terabyte hard drive (yet!) so you'd need to start pairing drives together, generally via RAID, to give yourself a larger accessible volume to throw that 5 terabyte file onto. An example might be an XServe RAID, which as of now can give you about 5.6 terabytes for $13k.
Of course, what if you are dealing with something that is 7 terabytes? Then you can take some XServe RAIDs and slide them together for an even larger data pool, all going through a type of abstraction, like Apple's upcoming XSan software. It can sound pie-in-the-sky, especially when you make it sound like one file, but there are situations where this occurs with larger sets of media or in a more networked setting.
Right now, this isn't really something we deal with on a consumer level, but sometime in the future it will be. IE, you'll stop caring what drive whatever file you want is on, it'll just be on the network and an abstraction will put it wherever is most appropriate from a data pool you've created from all the storage available, no matter on what machine or where. Cool stuff.
Sorry to ask a stupid question, but what is a data pool?