Database Reference
In-Depth Information
When you write all the new rows sequentially, they all end up being on the same
server, because they are sorted and this forces them to be close to each other. It is true
that there is sharding, and moreover, there is automatic sharding. This means that
the new regions (the areas of the hard drive where the data is written) will eventually
come into play, but this happens only later. Right now, you have got a hotspot.
Practically, you won't notice this under low write speeds. If you have less than a
hundred writes per second, this is not important since your RegionServer copes quite
all right. This number (hundred writes per second) might change on your hardware
and HBase, but it will be in the low range, because it means that you are not utilizing
your entire HBase cluster but only one server. This is how Ikai illustrates this:
