Hi!
And what is the scheduling policy of DHT? Is there something like NUMA planned?
Best regards,
Andrey
Andrey,
I think the DHT will distribute things equaly for all the subvolumes. It does that according with a hash of the filename. It doesn`t use weights on subvolumes, although it should be easy to implement.
This is why the DHT is great, the lookup() and open() operations will go only for the only node that have the file. DHT is also great because it doesn't have the namespace cache, but I think unify will remove this too...
Tomorrow I`ll get the new storages and start the tests.
Best,
Daniel
