Messages in this thread | | | Date | Thu, 1 Oct 1998 13:46:15 +0100 | From | "Stephen C. Tweedie" <> | Subject | Re: Building Big Ass Linux Machine, what are the limits? |
| |
Hi,
On Wed, 30 Sep 1998 14:08:27 -0400 (EDT), "Adam D. Bradley" <artdodge@cs.bu.edu> said:
> Actually, "snapshot" functionality is remarkably easy to do with an LFS, > and I imagine it's not that difficult for a JFS either. Both have the > advantage that old data can be pushed off to tape or other slow media once > it's no longer "live", but supporting this well would probably require > userspace help.
Snapshots are almost trivially easy with LFS. On the other hand, having a journaled filesystem is absolutely no help at all for constructing snapshots: journals are short-term activity logs, and snapshots need long term records of multiple versions of the filesystem.
Journaling does not help with migrating data to tape, either. Those are completely separate issues.
Right now, the only Linux filesystem that could be modified to support snapshots without a complete redesign is the tree-based Reiserfs: if you can support multiple root blocks representing distinct but overlapping filesystems, then you have everything necessary for snapshots.
One thing that we CAN conceivably do with ext2fs is to implement multiple versions for files (the redesign of directories for btree support will make this relatively simple). That is very, very different from snapshots, though. Supporting data migration to tape is a completely different issue, and we can do that without using any logging or journaling features.
--Stephen
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |