The Next Wave

Thursday July 13, 2006

The disposable enterprise

Just an interesting observation ... Sun just announced a bunch of new boxes, one of my favorites being Thumper, a four-way x64 server with up to 24 terabytes of storage. Finally, something big enough to allow me to packrat just about everything ... :-)

Seriously, it's an interesting idea where by using next generation file system software like ZFS, you can treat everything you have as just one big volume ... and if a disk drive fails, no big deal ... it becomes the equivalent of a marked off bad cluster on the single drive in your PC. You don't worry about replacing the drive (that's down time), just have it fail in place and leave it there. At some point when you add additional storage and migrate that data off, perhaps then you'll replace it. Given MTBF that may be a while ... so you have what is in essence disposable storage ...

The way things are going, we're fast approaching (and already there in some cases) the point where the computing environment itself is just as replaceable ... if a node goes down, processing just shifts somewhere else. You don't bother to replace or repair that node ... at least not until the impact is felt on the grid, cluster or network that it belongs to. Sending in a repairman just introduces risk ... fail in place is the way to go.

In terms of software of course this means we need to design things differently. Locality shouldn't and can't matter and you see that in the nature of current and next generation software platforms and technologies, where resources (compute, data or otherwise) are allocated from system-wide pools rather than fixed-node ones.

Brings new meaning to the disposable razor (blade that is) ...

Posted by brewin Jul 13 2006, 01:28:41 PM PDT Permalink Comments [1]