minimond is a "minimalist" approach to metrics collection, designed to
collect large numbers of metrics with minimal performance impact to
the system on which it runs.  Any calculation of derived statistics is
intended to be performed on a remote collection/analysis server.


The code currently does the following:
   1) Runs several input modules that collect information from a
      few files in /proc/ and stores them in memory
   2) Runs one or more output modules to display the metrics or
      send them elsewhere


Currently, the only output modules are "text" (which prints the
output to a logfile) and "gmetric" (which uses the "gmetric"
utility to send the data to Ganglia).  Additional output modules
for EmbeddedGmetric and AMQP could probably be written fairly
quickly.


For the prototype, I chose ease of implementation as the
standard for appropriate design decisions, so the code has the
following known issues/limitations:

   * Some memory is wasted due to statically sized data
     structures and buffers that may be larger than needed.

   * There is currently no support for calculating a change in
     metric values (e.g., bits per second).

   * There is no documentation.

   * Some overall cleanup is needed.


Please see the INSTALL file for installation instructions.