IBM Planning 120PB Cluster Array.

[Chewy509]

Saw this linked elsewhere, but:

http://www.extremetech.com/computing/94082-ibm-builds-120-petabyte-cluster-made-out-of-200000-hard-drives

Now that's a lot of pr0n... ;)

[Handruin]

Damn...the noise and vibrations must be intense.

[ddrueding]

Several Terabytes per second. Sweet. I wonder what kind of SSD cache they have strapped on the front end?

[Handruin]

It sounded like those transfer rates were a benefit of their custom filesystem. There is certainly that issue with the more data you store, the faster you need to read and write to make it worth accessing.

[ddrueding]

Indeed. If you connected a single SATA 3.0 link to 120PB, it would take 5.3 years to fill it.

[LunarMist]

I wonder what hard drives they are using?

[time]

Several Terabytes per second. Sweet. I wonder what kind of SSD cache they have strapped on the front end?

Why wouldn't they just use DRAM as normal?

[time]

I wonder what hard drives they are using?

2.5" form factor for starters. Aren't Seagate and Hitachi the main enterprise drive manufacturers?

[LunarMist]

Then they will be WD-Hitachi or the Seagate drives?

[LunarMist]

Why wouldn't they just use DRAM as normal?

Maybe DD expects OCZ. :D

[ddrueding]

Considering everything in that kind of rig needs to be redundant anyway, just get the fastest you can and let 'er rip.

[Chewy509]

just get the fastest you can and let 'er rip.
Except when 30% of the drives decide to fail at the exact same moment, and you're left with a slow-a*s array due to all the rebuilds taking place.