Hacker News new | past | comments | ask | show | jobs | submit login

I really wonder how this compares to the Tahoe LAFS (least authority file system) which was built for similar reasons and has been around a lot longer:

https://tahoe-lafs.org/trac/tahoe-lafs

And in specific:

https://tahoe-lafs.org/trac/tahoe-lafs/browser/trunk/docs/ab...

Or the Redhat sponsored HekaFS:

https://fedoraproject.org/wiki/Features/HekaFS?rd=Features/C...




I think Ori is more similar to Dropbox or BitTorrent Sync in the sense that it focuses on creating (nearly) full replicas on a fairly small number of machines with P2P sync between them. The Ori paper explicitly says disk space is cheap, which motivates their decision to store full replicas and history. It's really a quite different design than client-server cluster filesystems like Tahoe, Gluster/HekaFS, Ceph, HDFS, pNFS, etc.

I think Ori is confusing people by calling itself a file system instead of a sync/backup system. Sure, it's implemented using FUSE, but it doesn't behave like a traditional network file system.


I think it's OK being a file system.

All a fs is really is a specialized key-value store. Redis is a file system.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: