Skip to main content
Back to Projects

Stanford Living Heart Project — HPC for Cardiac Simulation

atNorth / UberCloud2017-2018HPC Infrastructure Lead
HPCCFDResearchStanfordUberCloudInfiniBandSimulation

Overview

The Living Heart Project is a collaborative initiative between Stanford University, Dassault Systèmes, and UberCloud to create a biophysically detailed, functional model of the human heart using finite-element simulation. The goal: a validated digital twin of the heart that could accelerate cardiovascular research and enable personalized treatment planning.

HPCFLOW (via UberCloud Experiment) provided the HPC cloud infrastructure that made the large-scale simulations possible.

The Challenge

Finite-element models of the human heart are computationally intensive — the Living Heart model contains over 600,000 elements representing cardiac tissue, valves, blood flow, and electrophysiology. Running these simulations at meaningful resolution required:

  • Bare metal compute: Virtualization overhead was unacceptable for tightly-coupled MPI workloads
  • High-bandwidth, low-latency interconnect: InfiniBand for MPI communication between nodes
  • Large parallel jobs: Multi-node runs that required reliable fabric isolation
  • On-demand access: Research teams needed to iterate quickly without managing infrastructure

What HPCFLOW Provided

  • Bare metal HPC nodes with InfiniBand interconnect via the HPCFLOW platform
  • Slurm-based workload management for job scheduling and resource allocation
  • Multi-tenant network isolation ensuring research workloads ran in dedicated fabric partitions
  • HPC expertise and hardware utilization guidance to optimize cluster configurations for the simulation workloads

Recognition

The Stanford/UberCloud Living Heart collaboration won Cloud HPC Awards from three organizations at SC17 (Supercomputing 2017):

  • Intel HPC Innovation Excellence Award
  • HPCWire Editors' Choice Award
  • Hyperion Research Innovation Excellence Award

Related Work

During the same period, HPCFLOW infrastructure supported UberCloud Experiment #200 — Human Brain Project: HPC infrastructure for personalized non-invasive clinical treatment simulations of schizophrenia and Parkinson's disease, published in HPCWire (October 2018).

Both projects demonstrated HPCFLOW's capability to support rigorous academic research requiring reproducible, high-performance compute at scale — not just commercial workloads.

Technologies

  • HPCFLOW bare metal provisioning platform
  • InfiniBand high-performance interconnect
  • Slurm workload manager
  • MPI (parallel computing framework)
  • Finite-element simulation workloads (Abaqus/FEA)