 |
 |
 |
 |
FLOW-3D/MP: CFD High Performance Computing
FLOW-3D users with access to Linux or Windows Compute clusters can obtain the same highly-accurate results for their transient free-surface problems by using FLOW-3D/MP, a high performance computing software for CFD. The latest release, Version 4.0, enables engineers to take advantage of the scaling potential of distributed memory computing on 64-bit clusters.
Achieve 16x Runtime Improvements with FLOW-3D/MP v4.0
FLOW-3D/MP Version 4.0 offers users substantial performance improvements, both in terms of simulation times and ease of setting up problems. Crucial developments to the software have reduced the amount of data sent via MPI communication, cut down on the number of synchronization points and improved computational load balancing using the automatic domain decomposition tool (ADT). ADT allows users to decompose the domain quickly and efficiently, instead of relying on a tedious and often inefficient manual procedure.
All of this makes it possible to extend scaling to up to 32 cores and achieve runtime improvements of as much as 16x. FLOW-3D users typically tackle transient free-surface problems, which are a difficult challenge for distributed-memory computing. Nevertheless, with FLOW-3D/MP Version 4.0, problems which previously took a week to complete can now be solved in hours.
The physical models and numerical methods of FLOW-3D/MP v4.0 are based on FLOW-3D v9.3.2 (pdf). Read more about the new developments in FLOW-3D/MP in our Fall 2009 Newsletter.
FLOW-3D/MP Benchmark Data
This report details eight benchmark tests run using the distributed-memory version of Flow Science’s computational fluid dynamics software, FLOW‐3D/MP, version 4.0. All benchmark tests were run on the same hardware configuration, using a standard procedure. The tests are representative of the type of transient, free‐surface flow calculations which FLOW-3D users tackle daily. The results of the testing show that, with FLOW‐3D/MP, users can achieve a speed‐up of as much as 16x on 32 processors and as much as 10x on 16 processors.
Download FLOW-3D/MP Benchmark Data >
FLOW-3D/MP demonstrates significant performance improvement for two representative simulations
FLOW-3D/MP v4.0 Certified by Intel as Cluster Ready
High performance clusters offer significant performance gains over typical desktop computers but can be quite complex to configure. There are many choices for operating systems, middleware, interconnects, memory, and storage. Intel Corporation has introduced the Intel Cluster Ready certification program to assure users that hardware and software will work correctly together. Hardware that displays the Intel Cluster Ready logo is assured to run software which is also certified as Cluster Ready.
Software and Hardware Requirements
FLOW-3D/MP's efficient
multi-block meshing at work.
FLOW-3D/MP runs on both SMP machines and clusters (which may consist of a group of similarly configured Xeons®, Pentiums®, Athlons® or Opterons®) running Redhat Enterprise 3 or 4 or Suse Enterprise 9 or 10, or Windows Compute Cluster Server 2003, a high speed network interconnect such as Gigabit Ethernet, or Infiniband and a large shared NFS disk accessible from all nodes in cluster. Prior to installing FLOW-3D/MP, a supported MPI library software should be installed. MPICH2 1.1.p1 is supported on both Windows and Linux platforms. In addition, Intel-MPI 3.2 is supported on Linux platforms and MS-MPI (part of Microsoft’s compute cluster software) is supported on Windows.
FLOW-3D users who do not have access to Linux or Windows Compute clusters can still get faster solutions with FLOW-3D by upgrading to the SMP parallel version for workstations with two or more processors or cores. Depending on the type of simulation and hardware in use, a speed-up of from 25% to 80% on two processors or cores can be realized.