Association for Computing Machinery
Rebalancing policies, which replace instances in lower preferred clouds with instances in higher-preferred clouds, are needed to meet these preferences. The authors present an environment that manages multi-cloud deployment rebalancing by terminating instances, in lower-preferred clouds and launching replacement instances in higher-preferred clouds to satisfy user preferences. In particular, users define a preferred cloud ratio, e.g., 75% instances on one cloud and 25% instances on another, which they attempt to achieve using rebalancing policies. They consider three rebalancing policies: only idle excess instances are terminated, excess instances are terminated gracefully, and worker instances are aggressively terminated, even if they are running user jobs.