Users pay to use resources in cloud systems which makes them more demanding on performance and costs. Optimizing the response time of the applications and meeting user's budget needs are therefore critical requirements when scheduling applications. The approach presented in this work is a scheduling based-HEFT algorithm, which aims to optimize the makespan of tasks workow that is constrained by the budget. For this, we propose a new budget distribution strategy named Estimated task budget that we integrate in our budget-aware HEFT algorithm. We use a multiple datacenters cloud as a real platform model, where data transfer costs are considered. The results obtained by our algorithm relative to recent work, show an improvement of makespan in the case of a restricted budget, without exceeding the given budget.