The SGD classifier software will use all the cores for training even without Hadoop.
Hadoop can definitely run on a multi-core machine, but the overhead introduced will mean that your net gain will be distinctly less than 8x. On Sat, Oct 2, 2010 at 6:43 PM, Latency Buster <[email protected]>wrote: > >If you have a multicore machine you may see performance improvements using > Hadoop >even on a single box. > I tried using Hadoop but was unsuccessful as I ran into installation > errors.. That was though, 3-4 months ago. But I think I need to > revisit this as I now have a Nehalem machine with 8 cores.. I probably > can extract better performance out of it if I use Hadoop. > > BTW: Is it recommendable to run Hadoop on a multi-core machine? It's > really not a cluster in that sense.. >
