One advantage of this approach is that if you're already using S3 (as in the SlateDB case mentioned in the article), it's essentially “free”. And it means that a package like SlateDB just needs to be pointed at an S3 bucket, instead of making you point it to an S3 bucket _and_ a DynamoDB instance.
Provisioning a DynamoDB table is scarcely more effort than provisioning a bucket. And you get way nicer constructs (still not as nice as RDBMS) for locking like ConditionalUpdate and TransactWrite.
Totally. I'm not advocating that people don't ever use it for this. I'm just saying that from a pure resource perspective, it might be one of the least resource-efficient mechanisms for doing this.
As others have pointed out, it's probably not a noticeable cost and, in fact, the fixed costs associated with setting something up yourself would far outweigh what you're paying to use S3 for this purpose.
Part of me just dies inside when I think of all the stuff needlessly happening behind the scenes, given it's not actually being used for storage. I mean, it's called Simple Storage Service.
I'll add that if a lot of people actually did start using it for this purpose, they could probably just productionize the thing they actually use for this, which is essentially their own version of etcd, but with Paxos instead of Raft.
