On 18/04/2014 19:06, Randy Black wrote:
The end goal would be to be able to pull data from the cluster without
having to go through a file system interface.
You are looking for a filesystem provides a direct or nearly direct
mapping of the object store.
This is a valid thing to want, but it's not what S3QL aims to do.
S3QL provides a filesystem layer which provides filesystem semantics
similar to traditional unix-style filesystems.
For example, with S3QL you can
* Rename a file or directory with almost no cost. A direct-map
filesystem might need to copy-and-delete thousands of objects to achieve
the same result
* Create symbolic and hard links. No practical way of implementing in a
direct-map filesystem
* Update a portion of a very large file without having to re-upload the
whole file. A direct-map filesystem will usually need to re-upload the
entire file
* Compress data before uploading. A direct-map filesystem will not be
able to do this, by definition.
* Deduplicate data. No practical way of implementing in a direct-map
filesystem
The "price to pay" is that you have to use the filesystem driver to
access the filesystem.
Regards,
Bill
--
You received this message because you are subscribed to the Google Groups
"s3ql" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
