On Fri, Aug 31, 2012 at 7:55 AM, Mohit Anchlia <[email protected]> wrote: >> My data is timeseries and to get random distribution and still have the > keys in the same region for a user I am thinking of using > md5(userid)+reversetimestamp as a row key. But with this type of key how > can one do pre-splits? I have 30 nodes. >
If you don't know the key spread ahead of time, let HBase do the splitting for you? St.Ack
