OpenStack Liberty Design Summit

Currently OpenStack does little to help Telcos optimize their workloads for energy consumption, cost, and speed.  Today, operators must manually (or via scripts) provision, migrate, and decommission workloads to achieve the desired balance of energy/cost/speed, and they must do so repeatedly.  

In this talk, we describe an open architecture for automating resource optimization, where operators provide a policy describing how workloads ought to be optimized, and OpenStack continually monitors and migrates workloads to satisfy that policy.  Under this architecture, operators give their policy to Congress [1] (the not-yet-incubated OpenStack project for Policy-as-a-Service), and Congress continually enforces that policy by migrating workloads as appropriate.  In addition to discussing the architecture, we demo a proof-of-concept implementation where Congress migrates real VMs via Nova in response to changes in datacenter readings reported by Ceilometer. 

References:

[1] "OpenStack Congress," https://wiki.openstack.org/wiki/Congress