-Re: Chaining multiple reducers most efficiently
Binglin Chang 2012-03-04, 15:00
No... Maybe in the future
On Sun, Mar 4, 2012 at 10:23 PM, IGZ Nick <[EMAIL PROTECTED]> wrote:
> I would like to chain multiple reducers without an intervening map phase.
> I looked at ChainReducer but it seems to support chains of the for M+RM*.
> What I am looking for is M+R*. What would be the mos t efficient way to do
> it. Of course, I could use an IdentityMapper in between, but I don;t want
> to introduce an unnecessary mapper. I want to start the next
> shuffle-sort-reduce cycle right after the first reducer does its job. Is it
> possible? I am using hadoop 0.20.2-append.