Multidimensional CFD has been widely used to improve the design of practical combustors. It is important to include a detailed chemical mechanism for an accurate representation of real combustion processes. However, this may lead to huge computational costs when the chemical source term is computed with conventional direct-integration (DI) method. DOLFA (Database for On-Line Function Approximation), a storage-retrieval scheme, has been developed to accelerate the chemistry calculation. This paper demonstrates the performance of DOLFA and systematically studies the effects of control parameters including error tolerance, scale factors, reduced tabulation variables, maximum database size and CleanDelete value. Numerical experiments show that the choices of these parameters greatly affect accuracy and performance.