Watcher

Watcher

Watcher is an OpenStack module that takes advantage of CEP and ML algorithms/metaheuristics to improve physical resources usage through VM placement improvement. This page is currently a work in progress.

Motivation

The aim of cloud providers is to maximize the utilization of their data-centers by efficiently executing their customer applications with a minimal cost.

From their perspective, the following metrics are necessary to compare optimization strategies for a given workload:

• the cost reduction
  • number of hardware resources utilized (storage servers, compute servers, network controllers, network bandwidth, ...)
  • energy consumption
  • maintenance & monitoring costs
  • licensing cost
• the generated revenue. If possible, manage overcommitment with a minimum number of SLA violations and induced penalties.

The watcher module offers an open optimization solution for helping cloud providers to better satisfy those objectives on an Openstack cluster.

The watcher module will be designed in order to provide a very simple interface for a cluster administrator, hiding the underlying complexity of metrics/events handling and optimization strategies.

Use cases

Detailed use cases can be viewed in the Google doc here: Watcher Use Cases.

Below we list use cases and the Watcher release at which we hope to support that use case. Please feel free to add additional use cases that you would like Watcher to support.

Use case	Watcher target release
UC0 : on-demand optimization of the cluster	alpha
UC1 : regular optimization of the cluster	TBD
UC2 : Order direct placement (optimize placement)	TBD
UC3 : Order planned placement (optimize placement and planification)	TBD

Architecture

Watcher has been designed to use the same modules and architecture of other OpenStack components, such as:

• oslo.db
• oslo.config
• pecan
• WSME
• stevedore

More detailed architecture info can be found here.