Cassandra's SSTables, i.e., data files, are compressed by default using LZ4 compression. The type of compression used is configurable and can be changed at any time. Michael Burman and I have been doing some testing with the GZIP compression (Cassandra's DeflateCompressor), and it looks like it produces anywhere between a 60% to 70% savings in space over LZ4. Deflate is not as fast as LZ4, and the Cassandra docs indicate that read performance is a bit better with LZ4. The space savings is substantial enough though that I think we should make GZIP/Deflate the default for the data table.