That's what I mean in my first reply. Basically, think of # of mappers or reducers of a job as the parallelism it requests (so many subtasks of mine can run in parallel). But think of # of mapper- and reducer-slots as maximum number of parallel computing resources available to serve requests from all clients (like # of processors in a computer). Now the number of mappers/reducers of a job that can run at any given time is obviously limited to available resources, regulated by job priority and scheduler ...
Michael ________________________________ From: Thejas Nair <[email protected]> To: [email protected] Sent: Thursday, September 15, 2011 5:59 PM Subject: Re: Num Reducers is Very Small! The number of reducers will not depend on the number of slots that get allocated. If the number of slots are less than the number of reducers, then some reducers will wait on slots to be available. -thejas. typed on a tiny virtual keyboard On Sep 15, 2011 4:36 PM, "Norbert Burger" <[email protected]> wrote:
