With the emergence of Just-in-Time manufacturing, production control mechanisms that react rapidly to actual occurrences of demand are gaining importance. Several pull type control mechanisms have been proposed to date, but it is usually difficult to quantify how good these mechanisms are, as well as understanding the structural properties that make them desirable. By using a two stage model and an optimal control framework, we study some of these issues here. Our framework permits quantifying the performance of classical mechanisms such as base stock and kanban and more complex mechanisms such as generalized and extended kanban. We also analyze the tradeoffs between single versus multiple control points and service level constraints on the backorders.