Abstract: A cloud computing environment in which a plurality of data centers are populated through the launching of resources in a controlled manner. Parameters such as current usage, future projected usage and a future maximum capacity can be used decide in which data center to launch a resource. A data center can be chosen so as to minimize a possibility that the data center becomes constrained at a future point in time. Selection is such that an equal percentage of future capacity is maintained across data centers.

Abstract: Capability based placement can be used for placing virtual machine instances on server computers (which can be configured to support one or more virtual machine slots) that are capable of running the instances in an efficient manner. For example, capability tags can be determined from the detailed characteristics (e.g., detailed hardware, software, and/or other characteristics) of the server computer and/or slots. For example, capability tags can indicate capabilities such as disk throughput, network bandwidth, database support, encryption support, video editing support, etc. Requests to launch virtual machine instances can be received and capability tags can be determined from the requests. Servers and/or slots that match the determined capability tags can be identified and used for launching the instances.