A method for optimizing provisioning workflows in cloud computing is provided. The method comprises determining a time coefficient for each resource type or aggregated resource types which define the installation time of a current workflow; accessing the current workflow and creating a tree topology based on the current workflow which defines serialized and parallelized provisioning steps; dividing the tree topology into independent sub-paths; determining a timing condition for an execution time of an optimized workflow; creating an optimized workflow template maintaining the timing condition by: selecting cheapest installation methods and cheapest resources; and reducing the parallelization of the tree topology; creating the optimized workflow by merging the optimized workflow template with data and the resource types of the current workflow.