Ask HN: Will you hope S3 become a real file system?
1) eventual consistency, do not known whether it's consistent or not, could result in wrong answer for analytics.
2) lacking of cheap/atomic rename, also symlink, it's painful to manage huge amount of data.
EFS is great, but can only be accessed from same VPC, also too expensive. Setuping up a DFS (GlusterFS/Ceph/MooseFS) in the cloud is also expensive and hard to manage.
ObjectiveFS looks great, after tried out, it still kind of eventual consistent from distributed system's point view(other nodes can't see the updates immediately). A distributed version of S3QL will be great.
Are you also hoping to have a solution that's as easy and scalable as S3, but also provide the same ACID guarantee as file system? If you know a solution like that, please leave a comment, thanks!