Is there a technical overview of how GlusterFS distributes files across
nodes, which is up-to-date for 3.2?
I have found quite a lot of older documentation, from which I see:
* "Unify"
- it needs a shared namespace brick?
- there is a choice of schedulers
* "DHT"
- can't find out much about this; no matches for "dht" in the wiki
- presumably uses a filename hash to decide the target brick?
- is this the same as the "distribute" translator described here?
http://www.gluster.org/community/documentation/index.php/GlusterFS_Concepts
But in the GlusterFS 3.2 Admin guide, all you get told is how to create a
distributed volume or a distributed replicated volume, not how it works
underneath.
In the source code both xlators/cluster/dht and xlators/cluster/unify exist.
dht has had recent commits, unify not much apart from copyright updates.
Also, there is a very old list of translators at
http://www.gluster.org/community/documentation/index.php/GlusterFS_Translators_v1.3
but it doesn't mention the "dht", "ha" or "map" translators which are in
the source.
Any pointers gratefully received. I am going to need to do a bit of an
internal sell and explanation on Gluster, and if there's anything current I
can use that would be great.
Thanks,
Brian.
