In situations, where you use same nodes as hypervisor AND storage nodes at the same time, up to 50% of network bandwidth can be saved with more control over the VM images locations. Imagine situation: distributed+replicated volume, 2 replicas node1, node2, node3, node4 vm1, vm2, vm3, vm4 node1 runs vm1 node2 runs vm2 node3 runs vm3 node4 runs vm4 It would be pretty convenient to have the possibility to set location of at least one of the replicas to the node, where given VM is running (or at least where it is running most of the time). The question is, how to implement it. I don't know the internals much, but I guess replicator brick would be the place to start - but then there is problem, how to give the control to the user and mainly, how to store the information. I'm not too sure if one file of volume-wide preferences is scalable enough (but I don't have deployments with hundreds of thousands of VM images - I'm dealing mostly with hundreds).
Distribute xlator makes the decision of where to put the data. Changing the component.
This has been made available as part of 882278. By enabling NUFA option (a variant of distribute), placing of data locally is possible. Marking the bug as a duplicate. *** This bug has been marked as a duplicate of bug 882278 ***