On 20.11.2012 01:15, Ted Miller wrote:
On 11/19/2012 12:12 PM, Les Mikesell wrote:
On Mon, Nov 19, 2012 at 7:48 AM, Ted Millertedlists@sbcglobal.net wrote:
If RHSS is not available or suitable, other suggestions welcome. I need a file system/server with:
- primary function is serving MP3 files for playback in a radio
station environment in Haiti * if the system goes down all your clients (listeners) know it * they know it NOW * they know how long it takes to get it back up
- High Availability as the primary concern
- ability to administrate via web interface or similar by
non-Linux-savvy IT staff.
- ability to grow file system from 2-3TB to 20-50TB by simply
adding disks and/or adding 'bricks'
- clients will all be Windows computers, so files accessible by
CIFS
- critical application is read-only
- prefer a system that would continue serving files even if the
network goes down (but have not found such a system yet for Windows clients).
Is it possible to change the application so it uses http to get content or uses a distributed database natively? Distributed failure-tolerant systems are a lot easier if you don't even try to provide filesystem semantics that require a lot of atomic operations.
Application is commercial, not changeable. It wants to see a local drive, if possible. Will tolerate (with warnings) a network share. Most of the critical operations are read-only (play back a file on the air).
Then it looks like GlusterFS is your only choice. Otherwise Openstack Swift might have worked, too, if it hadn't been HTTP only.