Date: Fri, 10 Nov 2006 23:42:13 -0600 From: Eric Anderson <anderson@centtech.com> To: Igor Sviridov <sia@nest.org> Cc: pjd@freebsd.org, freebsd-geom@freebsd.org Subject: Re: geom s3 class (gs3?) Message-ID: <455562B5.7@centtech.com> In-Reply-To: <20061110150957.A14717@seagull.nest.org> References: <20061110150957.A14717@seagull.nest.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 11/10/06 17:09, Igor Sviridov wrote: > hi, > > I guess most people here heard about Amazon S3 storage service: > http://aws.amazon.com/s3 > > I would love to have a Geom class which would use S3 as a back-end. > While S3 api is good one for Web applications, using filesystem > is often simpler and sometimes the only way. > > Obviously mapping filesystem calls to S3 API may be too slow for many > applications but this can be mitigated by adding local write cache/queue > on another geom provider (disk), or modifying gmirror to return write > calls once they're committed to local mirror component (and using it as > queue for remote mirror component stored on S3). > > There is in fact Linux prototype implementation of similar idea, though > on slightly different level - "s3fs" filesystem via FUSE: > > http://dev.extensibleforge.net/wiki/s3/fuse > > and closed-source open-spec JungleDisk: > http://www.jungledisk.com/ Amazon's S3 looks more like a filesystem than a GEOM-like block device. It could be implemented as a very rudimentary file system I suppose. Eric -- ------------------------------------------------------------------------ Eric Anderson Sr. Systems Administrator Centaur Technology Anything that works is better than anything that doesn't. ------------------------------------------------------------------------
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?455562B5.7>