Skip to Main Content U.S. Department of Energy
Computing Research

Modeling and Simulation

The modeling and simulation, or ModSim, capability centers on developing and applying predictive tools and methods to enable the exploration of large-scale current and future computing systems and workloads. ModSim techniques provide a quantitative understanding of the performance, power consumption, and reliability of systems and applications. The ModSim capability is applied to guide architecture design, assess the impact of emerging technologies, optimize complex workloads, and dynamically optimize performance and energy use. ModSim forms a foundational technology for effectively using upcoming complex extreme-scale systems.

Key Capabilities

  • Integrated Performance and Power Modeling
  • Full-scale Workload Modeling
  • Architectural Design Space Exploration
  • Performance and Power Assessment of Novel and Emerging Technologies
  • Performance Modeling and Analysis Tools
  • Workload Optimization

Significant Projects

CENATE Website


The Center for Advanced Technology Evaluation evaluates emerging technologies in relation to their use in future large-scale system architectures. A key capability within this context is performance modeling, where measurements and empirical analysis conducted on small-scale testbed systems can be extrapolated to scales that are as yet unavailable for measurement. Performance prediction allows architects to understand workload performance at large scales and forms a critical step in assessing technologies evaluated within CENATE.

Funding Source: DOE Office of Advanced Scientific Computing Research

Data Cyclone

Performance modeling combined with empirical analysis can be used to identify performance bottlenecks in large-scale system architectures. However, performance models also can take this analysis a step further and quantify how proposed, or hypothetical, changes to the system architecture will impact performance and alleviate bottlenecks. Similarly, through modeling, architectural components or technologies can be combined into a hypothetical system architecture that can be evaluated on real-world application workloads. The Data Cyclone project aims to understand the impact of the combination of two such technologies: high bandwidth memory implemented using three-dimensional chip stacking and fine-grained, contention-free networking.

Funding Source: DOE Office of Advanced Scientific Computing Research



The Exascale Computing Project. is tasked with deploying an exascale computing platform that delivers performance improvement by a factor of 50 over today’s largest-scale systems. Before they are deployed, applications aiming to use this unprecedented level of performance must ensure they are on target to deliver on the performance potential afforded by new system architectures. Application-specific performance modeling is vital toward developing application software in parallel with such system software and hardware.

Funding Source: DOE Office of Advanced Scientific Computing Research


Evaluating Execution Models uses performance modeling techniques to quantify the impact that the underlying execution model has on workload performance. As large-scale systems increase in architectural complexity, tried-and-true execution models, such as Communicating Sequential Processes (CSP), may prove insufficient. By using performance modeling and analysis methods with emerging execution models and runtime software, existing approaches can be examined to determine if they will suffice in the future.

Funding Source: DOE Office of Advanced Scientific Computing Research

IPPD Website


Large-scale distributed workflows are increasingly important as vehicles for scientific discovery. Workflows are designed to execute in a loosely synchronous manner on a connected set of distributed and heterogeneous resources, making it increasingly difficult to design, analyze, and implement on large scales. Integrated End-to-End Performance Prediction and Diagnosis for Extreme Scientific Workflows aims to develop performance modeling techniques for these large-scale workflows to guide optimization efforts and improve overall performance.

Funding Source: DOE Office of Advanced Scientific Computing Research

Computing Research

Research Areas