For this little project, we are going to use the Happybase Python package. Happybase uses HBase’s Thrift API.

For our test, we are going to create a namespace and a table in HBase. We will do this in the HBase shell. To make things simple, our table is going to have only one column family - data, and we are going to accept all defaults.

% hbase shell
hbase> create_namespace "sample_data"
hbase> create "sample_data:rfic", "data"

The data for this project was downloaded from on 10 February 2016.


You can insert directly into a table with the Table#put() function. However, I recommend using Batch#put() instead. When the number of records reaches the batch_size, Batch#send() will be called. See the section on benchmarks for timing data.

When you’re done, be sure to call Batch#send() manually, to flush any remaining records to the database.


I ran the program for several batch sizes and averaged the results. As you can see from the time results, we can only go so fast - a little above 2 seconds - before we cannot insert any faster.

Batch (n) Time (s)
1 10.227
100 3.533
1,000 2.091
2,000 2.141
5,000 2.081
10,000 2.190

The code