A 15 node cluster takes much longer to deploy than a 2 node cluster. Are nodes created sequentially? If not, why does it seem it takes longer the larger the cluster?
There is some additional overhead per node in a cluster stemming from us synchronously launching each EC2 instance and waiting until AWS gives us the EC2 instance metadata before launching the next EC2 instance. It should be the amount of time it takes for AWS to list each new EC2 instance as being in the 'Initializing' state. This should be a few seconds per node.
The forthcoming support for running ODAS on AWS AutoscalingGroups (ASGs) should minimize the per-node cost as ASGs spin up mosts in parallel.