Collaborative Research: CCRI: Planning-C: Enabling Computer Architecture Simulation as a Service

University Of Utah, Salt Lake City UT

Investigators

Abstract

Simulators are essential to computer architecture as they allow designers to validate design ideas without producing real chips on silicon, which is expensive and time-consuming. However, there are currently many issues slowing the advancement of computer architecture design through simulation: 1) Fragmentation of the ecosystem; simulator models developed in one framework cannot be used in another. 2) High barrier of entry; extensive training and powerful hardware are required, limiting participation, particularly from marginalized populations and institutes with limited resources, and 3) Neglect of usability and user experience; poor result interpretability hinders insights generated from simulations. The project's impacts are empowering the computer architecture research community with a one-stop, web-based, and cloud-powered computer architecture simulation environment to support more complex computer architecture simulations and to allow more researchers to participate in the advancement of computing system design. The project’s novelties are creating new computer architecture design principles that bridge technical and human aspects of usability across the computer architecture simulation ecosystem. This project delivers an online platform that allows users to conduct complex simulator-based computer architecture research without leaving the platform. Central to the platform is a Model Zoo for researchers to publish simulator models, fork others’ models, validate models, and use others’ models in their experiments. This project unifies the ecosystem by building a compatibility layer to connect simulator models developed in popular simulator frameworks such as gem5, SST, and MGPUSim. This project also enables computer architecture simulation as a service (CASaaS) as users can configure experiments, run simulations in the cloud, and monitor the running simulations in real-time. The platform also provides an integrated visualization tool for result analysis. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.