The key technological constraint for sensor networks is power which is dominated by the communication component of each node. While a multi-hop communication architecture greatly decreases deployment costs and increases versatility, it also causes the communication cost between two distant nodes to be prohibitively expensive. Our goal is to demonstrate how judicious placement of a few gateway nodes at strategically selected places can significantly reduce both maximal latency and communication costs for data aggregation resulting in prolonged network lifetime and enhanced network utility. We identify two main technical problems: gateway placement for minimizing communication delay and for minimum communication cost and establish their complexity. We have established efficient integer linear programming (ILP) formulations to optimally address these problems. Additionally, for both objectives we have developed a series of lower bound techniques and novel negative selection statistically-tuned heuristic algorithms for large instances which can not be efficiently solved by the optimal ILP formulation. We have evaluated the effectiveness of our techniques on a variety of network sizes, and statistically establish the importance of strategic gateways.