xen-orchestra/forecaster.md

Forecaster

This feature is planned for 2016.

Thanks to machine-learning, we should be able to analyze previous events (load evolution, RAM usage, etc.) and to forecast what would come next.

It could leads to:

• forecast report and alerts inside advanced reports
• taking automatic actions before bottlenecks happened
• match the resources needed for a predicted load
• green IT! By cutting useless nodes in quiet moment and boot them just before the load rise

Example

The system detects that every Monday, you got a huge CPU spike on some VMs. It could:

• report you this information
• scale up those VMs just before those spikes, and scale them down after