Ciaran Wills <[email protected]> writes: > Another feature which I couldn't figure out how to achieve is per-host > dependencies (particularly in array dependencies) - if one task is > going to generate a load of data for another task to consume then I > could write it all to the local tmp drive if I know the dependent task > will run on the same host rather than pushing it across the network.
It may be possible to do so with a load sensor, or an epilog, to set up an appropriate complex for the presence of the data locally (like the Hadoop integration, for which I don't have a useful URL to hand). _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
