Maximum parallelism only comes with tons of data. Splits will happen automatically if you have enough data, a region will normally split when it any of its family is bigger than 256MB by default. This question was answered quite of few times on this mailing list, try seaching for "MAX_FILESIZE".
J-D On Mon, Feb 8, 2010 at 7:53 PM, Naresh Rapolu <nrap...@purdue.edu> wrote: > Hello, > > To extract maximum parallelism and resource utilization when using > TableMapReduceUtil, I would ideally want my table to span all my region > servers. > But the "split" operation on my table is not always doing this, if my table > is small. It sometimes chooses the same region server to place the split > chunk. > > Is there any other way to achieve this "maximum parallelism" ? Can we > command HBase to distribute my table equally among all my regionservers ? > > Thanks, > Naresh Rapolu. >
