Dynamic load migration to conserve energy using VMWare

[Anonymous user]

Has anyone been discussing the concept of automatically scheduling the movement of VMs to consolidated groups of server hardware in a physical area during times of light usage (after hours?) in order to conserve power?

Ideally many servers could be moved to a smaller subset of physical servers during times of low utilization (and then automatically re-dispersed, or better yet "pre-dispersed" to a their regularly-assigned physical machines just prior to regular times of processing demand.

With further automation, certain parts of the data center could be physically shut down during those times when no VMs were running on those physical servers (this would also permit InRow cooling to be shut down as well as servers - either via intelligent switched PDUs or direct remote control of the servers using IPMI, etc.). Physical servers hardware could be restarted using wakeup on LAN capability.

During these after hours conditions, if the utilization of a particular server were to increase to a point where it required that it be migrated to its regularly-assigned server (for example, unscheduled after-hours user access), the system could automatically energize that physical server & adjacent cooling and then move the VM back to the consolidated group of physical servers once the utilization came back down.

In other words, the system would function in two modes based upon a schedule:

* optimized for performance.
* optimized for power conservation.

Just wondering if anyone else is thinking about these ideas...

- John