Cray Intros Urika-CS AI |
July 16, 2018
is accelerating and simplifying its customers’ path to value from
artificial intelligence (AI). The company is introducing the Cray Urika-CS
AI and Analytics software suite for the Cray CS series and new Cray
Accel AI reference configurations to help IT and AI teams.
A typical AI workflow spans data preparation using analytic tools
through to model development and model deployment using AI tools. It can
be time consuming and resource intensive to implement. The Cray Urika-CS
AI and Analytics suite and new Cray Accel AI reference configurations
join key tools and system configuration guidance to simplify the
implementation of AI. The Cray Urika-CS AI and Analytics suite makes the
most popular AI and analytics tools and frameworks readily available on
all GPU- and CPU-based Cray CS series systems, including TensorFlow™,
BigDL, Apache Spark™, Dask, Jupyter Notebook and TensorBoard.
The Cray Urika-CS AI and Analytics suite is optimized for and supported
on the Cray CS series cluster supercomputers and brings analytics and
artificial intelligence capabilities to existing HPC and emerging AI
environments. Combined with the Cray Urika-XC AI and Analytics suite for
the Cray XC series supercomputers, these powerful tools will now be
available across Cray’s portfolio of computing solutions.
“Today’s organizations are faced with a daunting task: to quickly
implement and integrate new applications that take advantage of the
latest advancements in Artificial Intelligence,” said Mike Leone, senior
analyst at ESG. “Successful AI application deployments require that
teams overcome numerous hurdles, including infrastructure complexity and
a constantly evolving technology landscape. There's an increasing demand
for a simplified way to make AI workflows more efficient and sustainable
by leveraging a purpose-built system that simultaneously addresses the
integration, performance, and support challenges brought on by AI.”
Additionally, Cray is introducing updated Cray Accel AI reference
configurations, which are designed to give IT teams the ability to
quickly meet the needs of AI teams implementing the AI workflow. The new
Cray Accel AI reference configurations bring together Cray CS series
systems, Cray® ClusterStor™ high-performance storage and Cray Urika-CS
AI and Analytics software to provide system configuration guidance to IT
organizations that are selecting and implementing infrastructure to meet
AI team requirements. Updated and expanded Cray Accel AI configurations
are available in both a “Prototype” and “Production” configuration.
“Today’s AI solutions are either too narrowly focused on deep learning
or too ad hoc in their design approaches to meet the needs of AI and IT
teams developing and delivering AI applications,” said Fred Kohout,
Cray’s senior vice president of products and chief marketing officer.
“Cray is taking the pain and guesswork out of AI deployments so
companies can get to business outcomes faster.”
Cray Urika-CS AI and Analytics Suite
The Cray Urika-CS AI and Analytics suite
addresses four key issues facing AI and IT teams:
Access to a range of tools for the AI workflow: Data preparation and
model development require different frameworks and tools. The Cray Urika-CS
suite includes popular data analytics environments (Apache Spark™ and
Dask), data science libraries for Python (Anaconda distribution),
machine and deep learning frameworks (TensorFlow and BigDL) and
interactive environments to visualize data science work (Jupyter
Notebook and TensorBoard).
Simplified and distributed model training
with TensorFlow: Setting up and using TensorFlow for distributed model
training can be laborious. The Cray Urika-CS suite includes the Cray
Distributed Training Framework – originally developed for Cray XC Series
supercomputers running Cray’s Urika-XC AI and Analytics suite –
simplifies and reduces the time associated with configuring and running
TensorFlow-based distributed neural network training and can reduce the
time required to train deep learning models by leveraging the
supercomputing infrastructure available on Cray CS series systems.
Reduced time and complexity for the setup
and maintenance of AI and Analytics environments: Setting up and
maintaining a consistent environment for users can be a time-consuming,
seemingly never-ending task. With the Cray Urika-CS suite, Cray provides
an integrated and tested suite of tools – ready to run on Cray CS series
systems – that eliminate the time required by IT teams to download,
install, test, debug and distribute myriad AI tools and frameworks to
Support for open-source tools and
frameworks: With the Cray Urika-CS suite, Cray is the single point of
contact for support and updates.
Cray Accel AI™ reference configurations
The new Cray Accel AI reference
configurations are designed to simplify the choices for IT teams
deploying infrastructure for the AI workflow:
configuration for small AI teams: a single-rack configuration that
includes a combination of Cray’s CS500 CPU nodes and CS-Storm 500NX GPU-accelerated
nodes for data preparation, analytics, machine learning and deep
learning with Cray ClusterStor L300F flash and L300N disk storage
solutions for performant cost-effective data pools. The Cray Urika-CS
suite provides the software to best utilize this heterogeneous compute
and hybrid storage configuration.
Production configuration for larger or
growing teams: A multi-rack configuration with separate racks for
CPU-based compute, GPU-based compute and storage that allows for
flexible scaling as teams grow and needs evolve.
The combination of systems and software has been designed so that IT
teams can enable their AI users to focus on insight and results rather
than system and tools.
The Cray Urika-CS AI and Analytics suite and updated Accel AI reference
configurations will be available in the third quarter of 2018.