The Terabox

I recently came across a pile of discarded hardware at work, waiting to be picked up and recycled. Amongst the trash were two of maybe 10 or 12 Teraboxes, which Dan Gruhl, Bruce Baumgart and I designed in early 2001.

The Terabox with eight hard drive slots
This brought back memories of the cool stuff we were working on in the early days of the WebFountain project. The Terabox was a rack-mountable, dual processor machine with eight exchangeable hard drive bays. We started with eight 140GB disks in a linux software RAID-5 array, yielding 1 Terabyte of storage, hence the name. We actually used velcro to attach an additional systems disk inside the case to have all eight bays for the storage array.

This was a huge improvement over one of our previous storage systems, which we setup to demo some of the initial WebFountain applications: a cluster of 20 machines, each with six 17GB SCSI disks, connected into one file system using GPFS, a high-performance super computing file system, yielding nearly 2TB of total storage. It may be hard to understand how vast a difference this made, but we went from four racks of hardware, to just two machines; from 120 hard drives to just 16, reducing potential hardware, networking and software failures significantly.

Some things strike me as pretty crazy:

1) You can buy a 750GB drive for just $219 at Fry's right now, and 1TB drives will be available soon. Heck - I have over 1TB of storage in my desktop computer right now.

2) The discarded machine has been in use until just recently - quite a lifetime for any piece of commodity hardware.

Bruce went on to work on the Petabox for the Internet Archive, pushing the boundaries of storage systems, while we focused on analyzing the Internet. It's amazing that with Google and Web Search being household names, most people take this technology for granted now. I am still in awe of any system that breaks the Petabyte boundary and I am grateful for having experienced this storage revolution first hand.