Compression algorithms allow to control the effort to compress the data
which impacts the compression time and compression ratio in no-linear
ways by setting the compression level:
- doubling the compression-time, only slightly improves the compression
  ratio,
- more time spent to compress increases the CPU utilization (and memory
  usage) itself and impacts the QPS of the server/client.

Another factor which has an impact on QPS is network latency. The time
necessary to transfer data can be decreased by increasing the
compression ratio of transferred data. In this way, the compression
level can be used to tradeoff cpu/network latency and make it selectable
by the user.

Goal
- Allow client and server operators to specify default
  compression-levels
- Allow client and server negotiate fast and safe compression-levels

Approved-by: Shiva Shankar <shiva.y.shankar@oracle.com>
RB:23433