Bigdata SQL: Data Compression

Most analytic workloads are I/O-bound. In order to make analytic workloads faster, one of the first things required is to reduce the I/O. Apart from data encoding, another technique to reduce I/O is compression. There are multiple compression algorithms to choose from.
However, in a distributed environment, compression has an issue: compression must be splittable, so that data chunks on each data node can be processed independently of data in other nodes in the cluster.However, systems such as Spark Succinct are being innovated to work with compressed data directly.
Splittable column: This indicates whether every compressed split of the file can be processed independently of the other splits.
Native Implementations: These are always preferred, owing to speed optimizations that leverage native machine-generated code.