SIM4LIFE » Framework » ARES
ARES

 Framework: ARES

High Performance Computing Auto-Scheduler & Control

Sim4Life offers high performance computing to enable the investigation of complex and realistic models. Multi-threaded execution for modeling, meshing, voxeling, and postprocessing enables parallel processing of heavy tasks without disturbing the workflow. A fully integrated centralized task manager efficiently manages all computationally intensive tasks on the local machine or in the cloud.

The Asynchronous Remote Server (ARES) automatically distributes any computationally intensive task to local or remote HPC resources (GPU/ CPU, cluster, cloud).

Sim4Life features the fastest graphics processing unit (GPU)-enabled EM-FDTD and US solvers (P-EM-FDTD & P-ACOUSTICS).

The MPI parallelization-based FEM solver optimally uses multi-core processors, clusters, and supercomputers for extreme performance.

A unified interface supports cloud computing on any of the major providers (e.g., Amazon or Google).

 

 Key Features - Networking

  • Fully integrated centralized task manager
  • All functionality (remote computing, HPC) seamlessly integrated into Sim4Life & Python framework
  • Parallel processing of computationally intensive tasks (e.g., meshing, simulation, postprocessing)

 

 

  • Remote execution via cloud (e.g., Amazon), localhost, GPU server, MPI cluster, p2p, etc.
  • Heavily multi-threaded execution for modeling, meshing, voxeling, and postprocessing

 

 

  • Queuing control instances, statistics
  • Solver control via Sim4Life GUI
  • Intelligent job submission
  • Job progress via a web browser (http based) on mobile devices, etc.

 

 

 

 Key Features - HPC

  • AXE GPU libraries
  • ZMT HPC/CUDA libraries
  • Cluster-MPI for Linux
  • Improved OpenMP parallelization for all of the above mentioned

 

Hardware Acceleration Solutions:
  • K20/x - NVIDIA Tesla K20/x
  • K40 - NVIDIA Tesla K40
  • Servers/Workstations with multiple GPU cards (K20, K20x & K40)

 

  • AXE MPI engine (for multi-CPU/multi-core distributed clusters)