The Structural Simulation Toolkit
Using the supercomputers of today to build the supercomputers of tomorrow
Navigation: Home | Downloads | Getting SST | Documentation | Support
Navigation: Home | Downloads | Getting SST | Documentation | Support
As the science of high-performance computing (HPC) evolves, there is a growing need to understand and quantify the performance and compositional value of emerging technologies. Modeling and simulation techniques are well positioned to serve this purpose. The Structural Simulation Toolkit (SST) is a parallel discrete event-driven simulation framework that provides tools to enable co-design of HPC systems – from application to architecture. Tutorial participants will be introduced to key facets of conducting reproducible simulations of HPC architectures and infrastructures. This tutorial will be split into two self-contained sections: one focused on system-level modeling with an emphasis on application performance analysis and one focused on node-level modeling with an emphasis on architectural research.
If you are new to SST, we recommend attending at least the “introduction” portion of the first tutorial prior to attending the second.
June 3, 2025, Afternoon
The scale of modern distributed systems makes full detailed simulation of relevant distributed applications impractical. SST provides two workload modeling environments that allow researchers to create motifs/skeletons that can be used to drive network simulations and as scaffolding to incorporate more detailed simulations. The Ember environment is the mainstay for workload modeling within SST and has underpinned many large-scale system studies using SST. Mercury is a new environment within SST, based on SST/macro, which allows an approach more closely resembling direct execution. In this section of the tutorial we will examine the use of both workload modeling environments in the context of system-scale network simulations utilizing the Merlin network components.
June 4, 2025, Morning
As novel compute architectures gain prominence in the evolving landscape of high-performance computing, tools that enable accurate modeling and simulation of these technologies are crucial for advancing both research and design. The balar GPU component, integrated into SST, provides scalable, trace- and execution-driven simulations of GPU systems by leveraging GPGPU-Sim. Similarly, Tactical Computing Labs’ (TCL) Rev RISC-V component and the native SST vanadis CPU simulator extend SST’s capabilities, allowing for detailed exploration of emerging CPU architectures. Together, these tools empower researchers to model and evaluate both CPU- and GPU-centric systems with unprecedented flexibility and accuracy. Participants will gain insights into the integration and use of balar for GPU performance modeling, as well as the Rev RISC-V component for node-level simulations.
We will update this page with instructions on where to access slides, code, and precompiled binaries for the hands-on exercises prior to the tutorial. In the meantime, explore SST’s general documentation.