This isn't worth doing unless we can figure out a way to keep the complexity
low but for some use cases it would be nice to specify an ordering of the nodes
such that tasks/jobs will tend to sent to specific nodes. This could be useful
for cases where we wish to keep computation close to data stored on local discs
or in local dbs.
This could be implemented by having the job create an ordered array of channels
on creation and attempt to submit to those channels before/at the same time as
the main channel. It could also be implemented to support issue #43 by having
logic executed to determine what node is likely to get a task after it comes
out of the main queuing channels.
Original issue reported on code.google.com by ryanbres...@gmail.com on 27 Jul 2012 at 11:52
Original issue reported on code.google.com by
ryanbres...@gmail.comon 27 Jul 2012 at 11:52