OpenStack-Neat-Ceilometer

This work is an extension to OpenStack Neat. It has been realized in the context of the Master's Thesis:
"A unified framework for resources monitoring and virtual machines migration in OpenStack".
We have integrated Ceilometer into Neat. Hosts are monitored by checking on samples of a new meter, defined from
a combination of host cpu utilization and host memory utilization.
When a threshold is crossed, an alarm triggers and a request is sent to an endpoint. Behind this endpoint there are two
services which handle the request to bring the alarmed host into a normal state.

The thesis consists in an extension to OpenStack Neat (this repository), and an extension to Ceilometer
(https://github.com/MisterPup/Ceilometer-Juno-Extension.git)

The following changes have been added to OpenStack Neat:

• The Data Collector have been substited by the Compute Agent of Ceilometer in the task of polling samples
• The Local Manager have been substituted by the Alarm Manager, a new component that receive requests when a Ceilometer Alarm goes into alarm state because an host have been found overloaded or underloaded. It selects the subset of VMs to migrate from the alarmed host
• The Global Manager have been integrated with Ceilometer
• A new version of the heuristic Best Fit Decreasing for solving the Bin Packing Problem have been created by taking into account the combined meter to avoid migrations to overloaded hosts
• System V scripts have been realized to control Alarm and Global Manager under Debian-based distros