CPU and RAM allocation is currently used to determine how instances are being distributed to compute nodes. The underlying algorithm used by the engine is shown below. The output of the algorithm is the compute allocation factor, which indicates how available a compute node is to run the instance; compute node with the lowest allocation factor will be selected to run the instance.

**Note:** Instances can be manually placed on a particular compute node by using compute categories and tags.

Greg DePasseIs the "Maximum available CPUs on compute node" in the last CAF equation before or after the instance is started?

Greg DePasseAnd how do the constraints factor into this?