[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ogfs-users]Cooperative backup using OpenGFS and iSCSI

While I'm sure this is possible with OpenGFS, I don't think it is the best 
solution for a setup like you describe. OpenGFS doesn't scale to a large 
number of nodes like you seem to think it could grow. You may want to take a 
peek at lustre. I hope this helps.

--Brian Jackson

On Saturday 27 September 2003 11:35 pm, Tracy R Reed wrote:
> I just read an article about iSCSI in Linux Magazine and it got me
> thinking. I have always wanted a distributed backup system where I could
> upload data to a cluster of machines scattered across the net and retrieve
> it if necessary. A machine would donate a certain amount of disk to the
> pool and that would entitle them access to a certain amount of storage
> depending on how much redundancy they wanted.
> Anyone see any potential problems with using OpenGFS over iSCSI to form a
> pool of disk on which each user would have their own logical volume spread
> across all of the hosts? They could use the cryptoapi and loop to encrypt
> the data they write into this disk so nobody else on the cluster can see
> their data. Let's assume for the moment that everyone involved are friends
> and are not looking to trash the system and we're probably only talking
> about a few gig per host.
> We wouldn't have STONITH capability or the ability to fence a renegade
> node out of the network with a fibrechannel switch which could be bad.
> Maybe everyone could automatically firewall off the offending system if
> something went wrong?
> If a node disappears/shuts down/crashes or otherwise becomes unreachable
> we would want to wait a certain amount of time before trying to rebuild
> the data onto another node so we don't initiate a rebuild just because
> someone reboots. Then if a rebuild does become necessary who would be
> transferring data where? Is there any way to make the cpu physicall
> attached to the disk download the data from the other systems and compute
> and write the info instead of doing multiple transfers over the network as
> another node does the math?
> Probably a crazy idea but I like crazy ideas.
> -- 
> Tracy Reed
> http://copilotconsulting.com

This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
Opengfs-users mailing list

[Index of Archives]     [Kernel]     [Security]     [Bugtraq]     [Photo]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Clusters]     [Linux RAID]     [Yosemite Hiking]     [Linux Resources]

Powered by Linux