Reliability, availability and maintainability (RAM) studies of oil and gas production facilities usually assess a design after it has been fixed, not choose it. This study embeds RAM analysis in design optimization for a representative onshore central production facility with seven series subsystems. Eleven decisions were optimized: redundancy and unit sizing for each subsystem, emergency shutdown voting, compressor and turbine spares, and condition-based maintenance. Production efficiency was computed analytically by a multi-state universal generating function, with spare stock-outs and spurious trips represented. All 15,552 designs were enumerated to obtain the exact Pareto front against life-cycle cost. Inputs are representative published values, not field data. The base design achieved 95.3 percent production efficiency, with compression responsible for 2.2 of its 4.7 points of loss. Condition-based maintenance gave the highest return per dollar, and a third compression train the largest absolute gain. A knee design reached 98.1 percent for USD 10.8 million of life-cycle cost. The net-present-value optimum reached 98.7 percent for USD 20.9 million and added USD 60 million. Spares were valuable alone but redundant once a third train was installed. Discrete-event simulation agreed with the analytic model within 0.1 percentage points. Under a factor-of-two uncertainty in failure rates, the optimum had a mean regret of USD 1.1 million, against USD 5.5 million for the knee design.