HPE Message Passing Interface (MPI): технические характеристики и документация
|
Have feedback on QuickSpecs? We're listening | |
|
HPE Message Passing Interface (MPI) QuickSpecs
An MPI environment designed to enable the development and optimization of high-performance computing applications The HPE Message Passing Interface (MPI) leverages a scalable MPI library and boosts performance of existing MPI applications on HPE HPC clusters without requiring recompilation. |
|
|
The HPE Message Passing Interface (MPI) accelerates application performance by: - Taking full advantage of the underlying server infrastructure. - Tuning applications at runtime without requiring recompiling the code - including applications developed with other MPI implementations. Restricting the number of processor and memory resources used for a specific process, or set of processes, to avoid oversubscribing the system and possible interference between applications. - Improving performance for I/O-intensive applications without recompiling or retooling the software logic. - Enable data placement to specific memory locations to limit communications overhead. - Providing access to tools and libraries for application profiling and analysis helping to identify performance bottlenecks and load imbalances as well as guided placements for threads. - HPE Message Passing Interface (MPI) v 1.9.7 | |
|
| |
|
What's New - Support for RHEL 9.5 and Rocky Linux 9.5 | |
|
| |
|
Models |
|
|
Licensing and Media Options |
|
|
HPE Performance Message Passing Interface for 1 Socket 3yr Support LTU |
Q2A48A |
|
Notes: - One license per socket. - Includes three (3) years of HPE Services support. - Please contact HPE Services for 4 and 5 year support options. | |
|
HPE Performance MPI FIO |
Q8K21A |
|
Notes: - For factory installation only; this SKU does not include the software license. Please order with PN Q2A48A. - Order one PN per node. | |
|
| |
|
Distribution Media and Software Documentation HPE Message Passing Interface (MPI) software is available for download. Upon receiving their order, customers will receive a physical document, which includes the website URL and additional license information. Software will be available for download via the Software Updates and Licensing Portal.
Patches may be required and will be delivered via the Software Download Repository. Release notes will be made available for any patches issued to customers.
Factory installation is available for the HPE SGI 8600, Apollo 20 and Apollo 40 systems. Additional platforms are not supported at this time. Please select PN Q8K21A for factory installation.
Customers may also download user guides and other documentation at www.HPE.com/software/mpi |
|
|
|
Software Specifications Standards and Programming Languages HPE Message Passing Interface (MPI) is compliant with these standards: - MPI-3.1 - OpenSHMEM 1.4
The Message Passing Interface (MPI) standard supports C, C++ and Fortran programs with a library and supporting commands. MPI operates through a technique known as "message passing," which is the use of library calls to request data delivery from one process to another, or between groups of processes. MPI also supports parallel file I/O and remote memory access (RMA).
In addition, HPE Message Passing Interface (MPI) supports the OpenSHMEM 1.4 standard. The OpenSHMEM standard describes a low-latency library that supports RMA on symmetric memory in parallel environments. The OpenSHMEM programming model is a partitioned global address space (PGAS) programming model that presents distributed processes with symmetric arrays that are accessible via PUT and GET operations.
Support for the MPI and OpenSHMEM standards is built on top of a message passing toolkit (MPT), which is available within the HPE Message Passing Interface (MPI). |
|
|
|
|
|
Operating System Compatibility The HPE Message Passing Interface (MPI) is operational on: - Red Hat Enterprise Linux (RHEL) 8.9, 8.10, 9.4 and 9.5 on x86_64 and aarch64 platforms - SUSE Linux Enterprise Server (SLES) 15 SP5 and SLES15 SP6 on x86_64 and aarch64 platforms - Rocky Linux 8.9, 8.10, 9.4 and 9.5 on x86_64 and aarch64 platforms - SELinux |
|
|
|
|
|
Supported Hardware The HPE Message Passing Interface (MPI) is supported on the following HPE systems: - HPE SGI 8600 - HPE Apollo Systems, including the Apollo 2000, 6000 and 6500 (Gen9, Gen10, Gen10+, and Gen11) - HPE Apollo 20 and 40 systems - ARM processor-based Apollo 70 and 80 systems - HPE ProLiant DL360, DL380, DL385 - HPE Integrity MC990X, HPE Superdome Flex family, and HPE Compute Scale-up Servers |
|
|
|
|
|
Fabric Support The HPE Message Passing Interface supports all major interconnects/ fabrics over multiple generations and does not require changes to software environment. Comprehensive fabric support and corresponding features make HPE Message Passing Interface (MPI) an ideal MPI library for HPE high-performance computing systems. Fabric support includes: - Multi-rail Intel® Omni-Path - Multi-rail Mellanox InfinibandTM Data transfer optimization for HPE-proprietary HPE Superdome Flex Grid where available, including single-copy data transfer TCP/IP |
|
|
|
|
|
Third Party Software Integration HPE Message Passing Interface (MPI) allows for tight integration with third-party debugging and profiling tools, as well as workload schedulers and fabric software. Such applications include:
Compilers and Debuggers: - Intel Parallel Studio XE* - PGI Professional Edition - Allinea Forge* - Rogue Wave TotalView* - Tuning and Analysis Utilities (TAU), Vampir - GCC, Gfortran, GDB
Workload Management Tools: - Altair PBS Professional* - SLURM - Univa Grid Engine - IBM Spectrum LSF - Adaptive Computing Moab/ TORQUE
Fabric Software: - Mellanox Fabric Collective Accelerator (FCA)
Notes: *3rd party software solutions sold by HPE. | |
|
| |
|
Interoperability with Other MPI Solutions HPE Message Passing Interface (MPI) can optimize performance of applications compiled with the following MPI implementations: - Intel® MPI - IBM Spectrum MPI - Open MPI - Mellanox HPC-X MPI - Cray MPI - MPICH - MVAPICH | |
|
| |
|
HPE Message Passing Interface (MPI) Features HPE Message Passing Interface (MPI) consists of the MPI library, message passing toolkit, and several complementary libraries and profiling and analysis tools that enable and speed development of scalable HPC applications. The Message Passing Toolkit (MPT) is a high-performance communications middleware software that is a component of the HPE Message Passing Interface (MPI). This toolkit is the core of the HPE Message Passing Interface (MPI) performance engine as it provides the flexibility users need to develop and run MPI programs. The MPT increases the speed of an application from launch. The environment includes HPE-proprietary software that enables parallel applications to run on multiple hosts in a cluster, resulting in the ability to launch 100,000 cores in a matter of seconds. The HPE MPI library accelerates performance of all HPE HPC systems and wide variety of interconnects (as per Specifications). HPE Message Passing Interface (MPI) also includes a tool that uses a wrapper library to run applications compiled against other MPI implementations through MPT, without requiring the user to relink or recompile their code. This allows users with applications compiled against Cray MPI, Intel® MPI, IBM Spectrum MPI, Open MPI, Mellanox X-MPI, MPICH, MVAPICH to still run their applications through the HPE Message Passing Interface (MPI) engine and increase overall application performance.
| |
|
Application Performance Optimization - HPE Message Passing Interface (MPI) accelerates application performance through tools designed to tune applications at runtime without recompiling code, as well as libraries which optimize performance through specialized algorithms. - These tools and libraries allow customers to: - Restrict the number of processor and memory resources used for a specific process, or set of processes to avoid oversubscribing the system and possible interference between applications. - Improve performance for I/O-intensive applications without recompiling or retooling the software logic. - Enable data placement to specific memory locations to limit communications overhead.
Users also gain access to profiling and analysis tools with HPE Message Passing Interface (MPI). One such tool captures communication analytics, pinpoints bottlenecks and identifies load imbalances during the application run. Once the tool is implemented and the application completes its run, users will receive a file outlining key statistics from the run. This information includes: - Time spent in computation, communication and file I/O - The size and number of each data request - The size of the communicator used for collectives - The number of times each rank was a root in a collective - Size distributions - Collective wait time and send late time
Users also have the option to analyze subsets of the application run by inserting start and stop commands into their program. The tool will begin analyzing the application run when the start command is executed and will generate the statistics file after the stop command is executed.
The profiling and analysis tools available within HPE Message Passing Interface (MPI) can be used for applications compiled against the Cray MPI, Intel® MPI, IBM Spectrum MPI, Open MPI, Mellanox X-MPI, MPICH and MVAPICH.
Additional features within the HPE Message Passing Interface (MPI) environment include a runtime tool to optimize processes and data placement on a set of nodes.
HPE Message Passing Interface (MPI) includes tools to optimize the placement of worker processes within the cluster. Users first run their application with HPE MPI's profiling tool to record the communication pattern between the workers. In future runs, the MPI's placement optimization tool takes this pattern information and combines it with information about where the application will be running to create a placement file assigning workers to specific locations in the cluster. This improves performance by putting workers that intercommunicate a lot close to each other, minimizing transfer costs.
It is important for HPE Message Passing Interface (MPI) fabric support to drive performance and resource optimizations, features include: - Adding more resources to a running application with MPI spawn is available for systems using Intel® Omni-Path, Mellanox® InfiniBand, TCP/IP, and HPE Superdome Flex Grid technology - Faster MPI performance for systems with NVIDIA® GPU with Mellanox® InfiniBand remote direct memory access (RDMA) - Increase application performance by creating fewer memory copies and non-blocking activity with Mellanox InfiniBandTM tag matching | |
|
| |
|
Software Licensing Information For the Software to be valid on an HPE Cluster, each socket in the HPE Cluster must have a valid HPE Message Passing Interface (MPI) license. Software licensing information can be found at https://www.HPE.com/us/en/software/licensing.html. Subject to the terms and conditions of this Agreement and the payment of any applicable license fee, Hewlett Packard Enterprise grants a non-exclusive, non-transferable license to use (as defined below), in object code form, one copy of the Software on one device (socket) at a time for internal business purposes, unless otherwise indicated above or in applicable Transaction Document(s). "Use" means to install, store, load, execute and display the Software in accordance with the Specifications.
Use of the Software is subject to these license terms and to the other restrictions specified by Hewlett Packard Enterprise in any other tangible or electronic documentation delivered or otherwise made available with or at the time of purchase of the Software, including license terms, warranty statements, Specifications, and "readme" or other informational files included in the Software itself. Such restrictions are hereby incorporated in this Agreement by reference.
Some Software may require license keys or contain other technical protection measures. HPE reserves the right to monitor compliance with Use restrictions remotely or otherwise. Hewlett Packard Enterprise may make a license management program available which records and reports license usage information. If so supplied, customer agrees to install and run such license management program beginning no later than one hundred and eighty (180) days from the date it is made available and continuing for the period that the Software is used.
Other terms of the HPE Software License are provided on the license agreement that is delivered with the HPE Message Passing Interface (MPI) software. |
|
|
|
For the most up-to-date information on HPE Services, please refer to the HPE Services - Supplemental QuickSpecs, which provides a comprehensive and regularly updated overview of available services. |
|
HPE Services Services |
|
|
Flex Support Services for 4 Year Support Plan |
SKU |
|
HPE 4Y NBD Proactive Care SVC |
H1K90A4 |
|
HPE 4Y NBD w DMR Proactive Care SVC |
H1K91A4 |
|
HPE 4yr Foundational Care 24x7 Service |
H7J34A4 |
|
HPE 4yr Foundational Care 24x7 w DMR SVC |
H7J35A4 |
|
HPE 4Y 4 hr 24x7 Proactive Care SVC |
H1K92A4 |
|
HPE 4Y 4 hr 24x7 DMR Proactive Care SVC |
H1K93A4 |
|
HPE 4yr Foundational Care CTR Service |
H7J36A4 |
|
HPE 4Y Foundational Care CTR DMR Service |
H7J37A4 |
|
HPE 4Y 6 hr CTR Proactive Care SVC |
H1K94A4 |
|
HPE 4Y 6 hr CTR w DMR Proactive Care SVC |
H1K95A4 |
|
|
|
|
Flex Support Services for 5 Year Support Plan |
|
|
HPE 5Yr Foundational Care NBD Service |
H7J32A5 |
|
HPE 5Y Foundational Care NBD DMR SVC |
H7J33A5 |
|
HPE 5Y NBD Proactive Care SVC |
H1K90A5 |
|
HPE 5yr Foundational Care 24x7 Service |
H7J34A5 |
|
HPE 5yr Foundational Care 24x7 w DMR SVC |
H7J35A5 |
|
HPE 5Y 4 hr 24x7 Proactive Care SVC |
H1K92A5 |
|
HPE 5Y 4 hr 24x7 DMR Proactive Care SVC |
H1K93A5 |
|
HPE 5yr Foundational Care CTR Service |
H7J36A5 |
|
HPE 5Y Foundational Care CTR DMR Service |
H7J37A5 |
|
HPE 5Y 6 hr CTR Proactive Care SVC |
H1K94A5 |
|
HPE 5Y 6 hr CTR w DMR Proactive Care SVC |
H1K95A5 |
|
| |
|
Date |
Version History |
Action |
Description of Change |
|
20-Jul-2026 |
Version 14 |
Changed |
- Updated GreenLake references to align with current branding and terminology standards. - Updated Supplemental Services QuickSpecs content for consistency and accuracy. |
|
16-Feb-2026 |
Version 13 |
Changed |
Visual rebranding only—updated typography, colors, and design elements to align with new HPE brand standards. No technical specifications or content were modified. |
|
28-Apr-2025 |
Version 12 |
Changed |
Overview and Standard Features sections were updated Obsolete SKU was removed; updated sections to reflect HPE MPI 1.9.7 release |
|
22-Jan-2024 |
Version 11 |
Changed |
Updated the following sections for MPI 1.94: Overview, What's New, Operating System Compatibility and Supported Hardware. |
|
04-Dec-2023 |
Version 10 |
Changed |
HPE Services Rebranding Obsolete SKUs were removed |
|
15-Nov-2021 |
Version 9 |
Changed |
Service and Support section was updated. |
|
09-Dec-2019 |
Version 8 |
Changed |
Overview and Standard Features sections were updated |
|
03-Jun-2019 |
Version 7 |
Changed |
Overview and Standard Features sections were updated |
|
17-Dec-2018 |
Version 6 |
Changed |
Overview and Standard Features sections were updated |
|
24-Oct-2018 |
Version 5 |
Changed |
Overview section was updated and PB ID was corrected in Summary of Changes section. Link was corrected in Standard Features section. |
|
04-Jun-2018 |
Version 4 |
Changed |
Updated the following sections: Overview, What's new, hardware compatibility |
|
04-Dec-2017 |
Version 3 |
Changed |
Updated the following sections: Overview, What's New, Models and Licenses, Hardware Compatibility |
|
25-Sept-2017 |
Version 2 |
Changed |
Updated the Models and Licenses section |
|
07-Aug-2017 |
Version 1 |
New |
Create QuickSpecs for HPE Performance Software - Message Passing Interface |
|
|
|
Have feedback on QuickSpecs? We're listening | |
|
|
|
| |
|
|
|
Chat now | |
|
|
|
© Copyright 2026 Hewlett Packard Enterprise Development LP. The information contained herein is subject to change without notice. The only warranties for Hewlett Packard Enterprise products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. Hewlett Packard Enterprise shall not be liable for technical or editorial errors or omissions contained herein.
Microsoft and Windows NT are US registered trademarks of Microsoft Corporation. Intel, the Intel logo, Xeon and Xeon Inside are trademarks of Intel Corporation in the U.S. and other countries. For hard drives, 1 GB = 1 billion bytes. Actual formatted capacity is less.
a00018446enw - 15997 - Worldwide - V14 - 20-July-2026 | |
|
|
|
HEWLETT PACKARD ENTERPRISE HPE.com |
|
