This dissertation focuses on two contributions in supply chain management when there is a risk of disruptions. The first contribution investigates the supply chain network design of a consumable product with risks imposed externally from the suppliers, while the second contribution concentrates on a rental product where the source of risk is internal. In the first contribution, we consider the supplier selection problem of a firm offering a single product via multiple warehouses. The warehouses face stationary, stochastic demand and replenish their inventory via multiple suppliers, to be determined from a set of candidates, with varying price, capacity, quality, and disruption characteristics. Additionally, the warehouses may simultaneously replenish their inventory from other warehouses proactively. We develop a decomposition based heuristic algorithm, powered with simulation to solve our problem. Experimental results show, contrary to the existing literature, inferior decisions may result when considering the selection of suppliers solely on unit and/or contractual costs. We also evaluate the impact of multi-sourcing with rare but long disruptions compared to frequent but short ones. In the second contribution, we discuss inventory decisions of a rental firm. Due to limited availability, some customers rent a product successfully, some may be backlogged, in which case they may choose to leave the system after a while, and some may be denied entering the system. Moreover, due to the usage by customers, a product may break down whether it is rented out or not and it becomes available to rent again only after repair or replacement. We model the problem as an M/M/c/c queue with customer reneging and product breakdown to find the best capacity. The current exact numerical methods to solve the model are either too time consuming or large accumulations of numerical error make them inapplicable as possible solution approaches. We develop a more efficient algorithm, exploiting a recursive form. Assuming the same reneging and return rates, our algorithm determines the global optimal capacity level. Additionally, our numerical study shows the problem is insensitive to the reneging rate, making our algorithm applicable to general situations. The dissertation ends with a discussion of extensions to each contribution.