Hi All, Trying to wrap my head around GlusterFS (pretty new to clusterd/distributed file storage systems) but am unsure on a few things that I can't seem to find the answers to.. Hopefully the members of the list don't mind clearing it up for me.. Firstly am I on the right track with my understanding of the storage methods.. This is how I read it and relate it to RAID terminology.. Distributed = JBOD Replicated = RAID1 Striped = RAID0 Distributed Striped = JDOD/RAID0?? Distributed Replicated = JBOD/RAID1?? If I had 10 bricks in a replicated volume would the the effective storage capacity be half the total storage capacity as is the case with RAID1? Next area I am not sure of.. Does GlusterFS only store complete files or chunks of files on bricks? I am looking to store Virtual Machine image files of anything up to 100GB.. Would the whole 100GB file be resident on one brick? or stored over a number of bricks? If running a replicated volume and a brick is removed permanently, will the data that now is not replicated be automatically replicated to another brick? Sorry for the simplistic questions but these things weren't obvious in the docs.. Thanks in advance.. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20120330/fbbbcec4/attachment.html>
>>>>> "Wipe" == Wipe Out <wipe_out at users.sourceforge.net> writes:Wipe> If I had 10 bricks in a replicated volume would the the Wipe> effective storage capacity be half the total storage Wipe> capacity as is the case with RAID1? It all depends on your setup: you can distribute, you can replicate, or a mixture. If you don't replicate, removing one brick will lead to loss. Obviously if you still have the disk, you can bring up a replicate of your lost brick with the original disks so you're back in business. Wipe> Next area I am not sure of.. Does GlusterFS only store Wipe> complete files or chunks of files on bricks? It can do both. If you replicate, it will store whole files. If you stripe it won't. See the manual. Wipe> If running a replicated volume and a brick is removed Wipe> permanently, will the data that now is not replicated be Wipe> automatically replicated to another brick? Not sure about this one, it definitely will if you attach a replacement. But it's not dynamic as far as I understand it. You specify beforehand how many bricks there are going to be, the system does not, or did not, handle a specification like: I want at least 3 copies of a file. It's more primitive, you specify I have 3 replicating bricks. But read the docs, they are actually pretty good. -- All the best, Berend de Boer ------------------------------------------------------ Awesome Drupal hosting: https://www.xplainhosting.com/
Sabuj Pattanayek
2012-Mar-31 00:02 UTC
[Gluster-users] Newbie's understanding of GlusterFS
> If I had 10 bricks in a replicated volume would the the effective storage > capacity be half the total storage capacity as is the case with RAID1?Yes.> > Next area I am not sure of.. Does GlusterFS only store complete files or > chunks of files on bricks?for stripes it stores a sparse file that shows the same "size" if you do an ls -l on the file not through the gluster layer, but it only has ~ fileSize/stripSize of the actual data (blocks) in the file per brick.> I am looking to store Virtual Machine image files of anything up to 100GB.. > Would the whole 100GB file be resident on one brick? or stored over a number > of bricks?if striped then stored over several bricks.