I wish I had the necessary petabytes of storage to at least store an offline copy. I wonder how many disks that would be and how redundant disks you’d need.
Here’s this from 2021. They say they have about 200PB of raw storage across some 20k spinning drives at the time of writing (with more being added constantly, about 25%/yr), and capacities are mixed from 4TB to 16TB, across 750 servers housed on about 75 racks. I have 6x16TB WD red pros that ran me about $355/ea new with tax, and my bill was a smidge over $2100. Assuming you used all 16TB, you’d need about 12,500 16TB disks, which would run you about $4,437,500 without a bulk discount. How much of that is redundancy I’m not sure, but that’s just HDDs, not the hardware to actually run everything between storage enclosures, OS, disks, memory, clustering, etc. They say they say a single copy with 16TB drives would be about 15 racks., but how that breaks down I’m not sure.
That’s unfortunately a very valid point. Iirc the big problem IA has is the sheer amount of disk space required to store everything.
I wish I had the necessary petabytes of storage to at least store an offline copy. I wonder how many disks that would be and how redundant disks you’d need.
Here’s this from 2021. They say they have about 200PB of raw storage across some 20k spinning drives at the time of writing (with more being added constantly, about 25%/yr), and capacities are mixed from 4TB to 16TB, across 750 servers housed on about 75 racks. I have 6x16TB WD red pros that ran me about $355/ea new with tax, and my bill was a smidge over $2100. Assuming you used all 16TB, you’d need about 12,500 16TB disks, which would run you about $4,437,500 without a bulk discount. How much of that is redundancy I’m not sure, but that’s just HDDs, not the hardware to actually run everything between storage enclosures, OS, disks, memory, clustering, etc. They say they say a single copy with 16TB drives would be about 15 racks., but how that breaks down I’m not sure.
I once made this calculation for a database of 700Tb, even that blew my mind 🤣