Open Stack

> We need to do some refactoring of periodic tasks anyway between
> nova-compute and nova-conductor.  I suspect that we can improve the
> situation as a part of that.

Indeed, although we have to be careful/aware of pushing something up to 
conductor that just makes the long-running request back to compute, 
which won't help anything. We also don't want to block conductor by 
running something long that will prevent it from answering database 
requests on behalf of the compute nodes :)

Maybe conductor would be a good place to start testing some real 
parallelism that could benefit the other event-driven-only services?

-- 
Dan Smith
IBM Linux Technology Center