> To avoid a DHT problem ;) it maybe better to take this sorted list and > assign tests in a cyclic fashion so that all chunks relatively take the > same amount of time to complete, than it being skewed due to the hash?
I assume you don't really mean cyclic, because that would *guarantee* a result in which the first worker takes longer than any other (because it gets the longest of every N tests), and so on. It's more of a bin filling thing, which is a fine idea but more complex to implement. _______________________________________________ Gluster-devel mailing list [email protected] http://www.gluster.org/mailman/listinfo/gluster-devel
