In this talk, I will introduce SMOOTHIE, a new framework for transaction processing. SMOOTHIE is based on optimistic concurrency control separated into a control flow of scalable components that can each scale. In addition, each of these components is nearly stateless and thus can be easily scaled up or down on-demand. I will present the architecture of SMOOTHIE, explain how it achieves scalability, present thoughts on how to avoid bottlenecks through heat-based data placement, and discuss tradeoffs between different implementations of the components of SMOOTHIE.