I’m pretty disappointed they didn’t mention this in the article. If this gets any traction it will eat as much CPU and bandwidth you give it.
Running with Docker and limiting CPU is necessary along with a bandwidth limiter/policer/shaper.
Having run IPFS for a long while it also ends up in strange states and stops responding to requests. I’ve taken to using a monitoring system to GET a known CID and restart the IPFS container if it fails.
My bandwidth and hardware have 100% uptime but IPFS needs restart at least a couple of times a day on average (looking at logs).
In short: this isn’t practical.