Abstract: Embodiments of the present invention generally describe how to utilize a pool of raw computing resources to dynamically resize environments due to workloads. Specifically, embodiments of the invention provide an approach for generating a partially pre-provisioned pool of virtual machines (VMs). The pool is typically derived from a history of patterns weighted by a potential state of a corresponding VM image to fulfill a workload request. For each VM instance, this approach determines/calculates the number of times that an instance represents a possible starting point for a VM request. The starting points are then normalized to create a probability distribution of the resulting partially pre-provisioned pool of virtual machines.

Abstract: Embodiments of the present invention provide a set of algorithmic methods that can be used to select which partially and/or pre-provisioned virtual machines (VMs) should be used as a base platform to satisfy a new workload (e.g., provisioning) request received in a networked computing environment (e.g., a cloud computing environment). Specifically, when a workload request is received, a set (e.g., at last one) of software programs needed to process the workload request is identified. Then, a set of VMs is selected from a pool of pre-provisioned VMs having the set of software programs. In general, multiple methods and/or factors can be followed to select the set of VMs. Examples include a length of time to install the set of VMs, a probability of the set of VMs being in demand, and/or or a quantity of the set of VMs having the set of software programs. Once the set of VMs has been selected, the set of VMs may be installed, and the workload request can be processed using the set of software programs.