System Modeling Challenges in a Late-Moore Environment

Making performance predictions and design decisions for future computer systems has long relied on models at multiple scales. In recent years, the declining impact of Moore's law has changed the role and impact of these models due to increased pressure to measure specific feature impacts and to improve the agility of decision making in a fast-changing application environment. How can models of the future adapt to this changing environment? What modeling technologies will help to measure the right things as quickly as possible?