This paper explains the benefits of using mathematical optimization to construct high performance reservoir operating rules and the related water rationing (deficit sharing) policies. The principal idea of the proposed approach is to generate perfect solutions obtained from an LP-based optimization model with the assumed foreknowledge of inflows represented with historical natural flows that are matched in the model with the current or projected levels of water demands. Water demands may include a mix of on-stream (e.g., e-flow targets or hydro power) and off-stream demands (irrigation or industry). The paper demonstrates the benefits of the proposed methodology by developing and testing short term operating rules on the Barna reservoir in Narmada River Basin in India. It shows that it is possible to achieve simulated results that follow the proposed rules and differ by only 2.5% in terms of the mean annual deficits from the best possible performance obtained using mathematical optimization with full foreknowledge of inflows.