Cloud computing offers customers an efficient way to flexibly allocate resources to meet demands. Cloud service vendors can offer consumers three purchasing plans, i.e., on-demand, spot, and reserved instances for resource provisioning. Since price of resources in reservation plan is generally cheaper than that in on-demand plan, in this study we attempt to make use of the cheap reserved instances to reduce monetary costs. We consider processing a large divisible load onto on-demand and reserved instances in clouds. Divisible loads, also called embarrassingly parallel workloads, can be partitioned into an arbitrarily large number of independent load fractions and be distributed across multiple processing nodes. We investigate the time-cost optimization problems for provisioning resources and scheduling divisible loads with reserved instances in clouds. The objectives are two-fold: First, given a total processing time (deadline), minimize the total cost. Second, given a budget (total cost), minimize the total processing time. We formulate the problems as mixed integer programs (MIP). We show that the optimal solutions of the problems have very simple structures. We then propose light-weight optimal solutions for the problems with rigorous proofs. Numerical experiments are presented to illustrate the salient features of these solutions.