> Also, I had a thought that a smart miner will just create a thin layer between Backblaze B2 and Filecoin to earn some FIL basically for free since no one is going to out-price B2. And then we're just back to storing everything on B2 and S3, except in a less efficient and more opaque way.
They can totally do this, but they are on the hook for that data. If the backing store for the data that the miner is trusting goes down, the miner loses a pretty significant amount of collateral (currently 0.2 FIL per 32GB sector).
In practice, we're seeing most miners set up their own storage clusters. Whether a bunch of small (100TB-1PB) nfs boxes, or much bigger single filesystem CEPH clusters.
I'm glad most of the network seems to be rolling their own solutions. Plus, what I said really only applies to lazy miners. It's probably not that hard to set up a thin compatibility layer between B2 and Filecoin, but as another commenter noted, they charge about $60/TB/year.
If you wanted to really be competitive on the Filecoin market, you could probably set up a rig just like Backblaze does (they open source their designs) and achieve the approx $30/TB/year mark, which is basically half of the B2 compat layer.
They can totally do this, but they are on the hook for that data. If the backing store for the data that the miner is trusting goes down, the miner loses a pretty significant amount of collateral (currently 0.2 FIL per 32GB sector).
In practice, we're seeing most miners set up their own storage clusters. Whether a bunch of small (100TB-1PB) nfs boxes, or much bigger single filesystem CEPH clusters.