Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
CRAY SHASTA SOFTWARE
NEXT-GENERATION SUPERCOMPUTING
SOFTWARE
SOFTWARE
SUPERCOMPUTING THAT RUNS LIKE THE CLOUDSOFTWARE FOR NEW WORKLOADSThe convergence of HPC, AI, and data analytics workloads coupled with explosive data growth have created new requirements for compute, software, networking, and storage. Future supercomputers will need to handle exabytes of data in order to enable modern workloads to run in a productive, reliable, and expedient manner.
Seeing the need for an evolution of supercomputing software infrastructure, Cray designed and developed the Cray Shasta software solution. Built on decades of supercomputing expertise, Shasta software adds the productivity of cloud and data center interoperability to the power of supercomputing to bring you a new standard in manageability, reliability, availability, and resiliency.
Cray Shasta software delivers the ease and flexibility of cloud computing through an environment which supports microservices-
based composability for rapid innovation of new converged workflows across processor architectures. Seamlessly scale from development in the cloud to production on a supercomputer.
With the Cray Shasta software stack you get Cray’s legendary package of HPC-optimized programming tools, Linux OS, and a scalable analytics/AI suite.
Cray’s Linux Environment is a suite of high-performance software designed to run large, complex applications and scale efficiently to more than 500,000 processor cores. Also, many ISV applications are available “out of the box,” benefiting from Cray’s network and MPI libraries.
The Cray Urika AI and analytics software suite unlocks sophisticated insights with a robust set of tools including Apache® Spark™, TensorFlow™, Cray Graph Engine, Python with distributed Dask, PyTorch, Cray Distributed AI plugin, Jupyter Notebooks and many more.
The Cray Shasta software stack can be broadly split into administrator and developer services.
• Systems, network, storage, and application management and monitoring support predictable performance of workflow
• Microservices approach with full separation of management plane from compute plane to allow each to be run and upgraded seamlessly ensuring high resiliency of critical system management software
• REST APIs & standard protocols enable full interoperability with existing monitoring, management, and automation toolsets
• Pre-integrated AI software packages that are current with evolving AI frameworks/tools and tools to Increase data scientist productivity
• Containerized developer environments enable portability from laptop or cloud to supercomputer
• Native support for containers and converged workflow orchestration on Cray’s Linux Environment delivers composability of cloud on hardened supercomputing OS
• Common software stack simplifies systems administration and support for converged scientific and AI workflows
Developer Productivity and Capability. Curated set of flexible and powerful tools to create new converged modeling, simulation, analytics, and AI workflows.
Value at Any Scale. Same features and capabilities through a unified software stack regardless of system size and spanning both liquid-cooled and air-cooled systems.
Operational Excellence. Core supercomputing capabilities combined with new cloud capabilities shifts focus from maintenance to innovation.
Datacenter Interoperability. Open REST APIs for management, access, and integration
Investment Protection. Complete software stack supports a broad range of operating systems and processor options from management to end user development
CRAY SHASTA SOFTWARE STACK
OP
EN
AP
Is
OP
EN
AP
Is
SOFTWARE
CRAY SHASTA SOFTWARE SPECIFICATIONS
©2019 Cray Inc. All rights reserved. www.cray.com, Cray and the Cray logo are registered trademarks, and Shasta is a trademark of Cray Inc. All other trademarks mentioned herein are the properties of their respective owners. 20191031WWW.CRAY.COM
OPERATING SYSTEM
CRAY’S LINUX ENVIRONMENT
• Compute node Linux optimized OS for efficient, low-noise parallel application execution• SUSE Linux Enterprise Server base
SYSTEM MANAGEMENT
CRAY SYSTEM MANAGEMENT
• Optimized for exascale-class supercomputing environments• Flexible, scale-out service mesh with micro-services architecture• RESTful API for automation and integration• Built-in monitoring at scale
SUPPORTED 3RD PARTY WORKLOAD MANAGERS Altair® PBS Professional®
COMPILER SUPPORT Cray Compiler, GNU Compiler, AMD AOCC, PGI Compiler (optional)
SCIENTIFIC LIBRARIES LAPACK, ScaLAPACK, BLAS, libsci, FFTW 3.0
COMMUNICATION LIBRARIES Cray MPI for Slingshot, Cray SHMEM
I/O LIBRARIES NETCDF, HDF5
DEBUGGERS GDB4HPC, Arm® Forge, Rogue Wave® TotalView™ debugger, STAT-Stack Trace Analysis Tool, ATP-Abnormal Termination Processing
SCOPING AND PORTING Reveal, CrayPAT Performance Analysis Tool, Apprentice2 visualization, PAPI
CRAY ANALYTICS/ AI SUITES
Analytic/AI Services - Open source and Cray optimized libraries and tools for Analytics and Distributed AI, e.g. Apache® Spark™, TensorFlow™, Cray Graph Engine, Python with distributed Dask, PyTorch, Cray Distributed AI plugin, Jupyter Notebooks.
Urika Manager – Tool to create and manage Analytics/AI container-based application runtime environment