Cost-benefit Analysis and Exploration of Cost-energy-performance Trade-offs in Scientific Computing Infrastructures

AbstractIn this work, we demonstrate how a cost-benefit analysis of different scenarios reveals insights into trade-offs between cost, performance, and energy for specific scientific computational work- loads. We leverage simulation to find a cluster configuration that will offer the lowest total cost of ownership for a scientific cluster. We also analyze at what point in time a hardware upgrade becomes beneficial, taking into account performance/watt improvements and hardware devaluation. This knowledge enables a person responsible for planning, designing, and managing the life cycle of a scientific computing infrastructure to find a cost-effective solution tailored to their specific applications and resource utilization patterns. We demonstrate that quite substantial cost savings can be achieved by using this methodology