20

BCR-2021-V1.1
 2021 State of Cloud HPC Report

Forew
ts and engineers, and they deserve the best we can provide.

We hope you enjoy this report.

Edward Hsu,
Chief Product Officer, Rescale

2
 2021 State of Cloud HPC Report

Big Compute
State of Cloud HPC Report

1 Hardware & Infrastructure Methodology

This report includes responses to a blind survey of 80 HPC practitioners

and analytics from the Rescale Cloud HPC Platform used by hundreds
2 Soft of customers and partnered with by leading cloud providers and 680+
independent software applications.

3 ment About Big Compute

4
Big Compute is a global community of thought leaders who embrace the
freedom to think big in this era of virtually unlimited compute. From an annual
conference of inspiring speakers, to telling stories year-round of the many
“undercover superheroes” working behind the scenes to change the world,
5 Big Compute fosters innovation in aerospace, automotive, AI, life sciences,
new energy, and everything in between.

For more information, visit https://bigcompute.org/.

3
 2021 State of Cloud HPC Report

High Performance Computing Powers Big Compute

Big Compute is the use of high performance computing (HPC) hardware and specialized Growth of Big Compute vs Big Data in 2021
softw
30.6%
better withstand collisions.
CAGR
Today, cloud represents the fastest growing segment for both Big Compute and Big Data. 24.6%
CAGR
2021
2021 Market Size:
Market Size:
$19.2B Cloud
$5.3B Cloud $64B Total
$42B Total

Big Compute Big Data

Source: MarketandMarkets and Hyperion Research

“I used to ask companies how they’re going to get lot of data, but now I ask them
how they’re going to get a lot of compute…. It turns out that the most impressive
advances that we’ve had in the field of AI research I think have been more about
massive compute than massive data.”
- Sam Altman @ Big Compute 2020

4
1
2021 State of Cloud HPC Report

Hardware & Infrastructure

Key Findings Spotlight: Nissan

• Cloud is seeing an explosion of specialized HW

architectures

• Multi-cloud becomes an operational reality for half of

enterprises

• Using the right architecture is often more impactful than Nissan leverages Rescale’s cost and performance
intelligence to select the optimal infrastructure based on cost
finding the ideal cloud provider and performance for their specific workloads. This modern
HPC approach ensures that they are always running the
latest, most performant hardware and software configuration.

Nissan

5
HARDWARE & INFRASTRUCTURE 2021 State of Cloud HPC Report

Growth of new architectures and enterprise adoption

accelerates
Cloud provider hardware options available Age of latest hardware adopted by organizations
300 40%
Total Cloud Provider Hardware Introduced

30%
32%

Percent of Organizations
200
28%
20%

100 18%
10%
11% 12%

0 0%
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 Less than 1 year old 1-2 years old 2-3 years old 3-4 years old 4+ years old

Year Age of Hardware

Cloud providers have accelerated the rollout of new HPC architectures Many organizations begin testing and tuning new architectures during the
with exponential growth. first year of introduction, but fully adopt the following year after optimal
configurations are determined.

Source: Rescale Cloud HPC Platform Analytics. Market survey of publicly available cloud provider architectures. 6
HARDWARE & INFRASTRUCTURE 2021 State of Cloud HPC Report

Ever more options and decisions

Cloud architecture specifications by year of introduction

An explosion of cloud-
3

Core Clock 2 based HPC architectural

Speed (GHz)
specs provides opportunity
1
to match every workload
with the ideal hardware.
10000

Finding the best SW/HW

1000
Core
Memory (MB) 100
match without getting
10

0
overwhelmed requires
500 workload performance
100
50
analytics.
Core I/O (Gb/s)
10
5

1
10 Coretypes 5 Coretypes 1 Coretype
2014 2015 2016 2017 2018 2019 2020

Source: Rescale Cloud HPC Platform Analytics 7

HARDWARE & INFRASTRUCTURE 2021 State of Cloud HPC Report

Performance vs fully-loaded costs for a sample benchmark workload (higher is better)

Recent cloud provider
25.00 adoption of HPC-
centric architectures has
20.00 DC HW refresh
contributed to a jump in
hardware performance.
15.00
Given the scale economics
Value

DC HW purchase

10.00
for admin costs long
Cloud provider
(top 4)
enjoyed by cloud providers,
it is increasingly difficult
5.00 Data Center
(fully loaded) for enterprise data center
operations to compete
0.00
2015 2016 2017 2018 2019 2020 purely on performance/cost.
Date available

Notes: Performance based on benchmark CFD workload with STAR-CCM+. Cloud: Reserved Instance list prices used
Datacenter: Assumes 100M core-hours / year consumption at 80% HW utilization and fully loaded costs for HW, labor and facilities. Server to FTE ratio of 50:1. Assumes median US data center labor costs
DC 2015 HW: Per node (6 cores) cost of $3K for Intel Core I7 processors; DC 2019 HW: Per node (18 cores) cost of $10K for Intel Xeon 6154 processors
Source: Rescale Cloud HPC Platform Analytics 8
HARDWARE & INFRASTRUCTURE 2021 State of Cloud HPC Report

Cloud HPC architecture shows signs of increased

diversity Cloud penetration of architectures used in HPC
While the first hardwares used in the cloud for HPC were
general purpose architectures, recently there has been Supercomputing architectures
increased shifts towards new architectures. Key HW
trends include:
HPC-optimized architectures

• Intel architecture usage remains dominant

• AMD usage growth 4x faster than Intel General purpose architectures

• ARM and related emerging architectures are

projected to make up ~30% of the server market
by 2025* 100%
0% 1%
Intel vs AMD Hardware 100%
9%
99%
• NVIDIA GPU architectures now available in the Usage in the Cloud 70%
91%

cloud see mainstream adoption in AI/ML applications

AMD Core Hours 50%

Intel Core Hours

As of printing of this report, the world’s fastest 20%

supercomputer Fugaku, developed jointly by RIKEN and 0%

2018 2019 2020
Fujitsu Limited, is based on Arm® technology. Core Hour Usage

Source: Rescale Cloud HPC Platform Analytics, Ark Invest Big Ideas 2021 9
HARDWARE & INFRASTRUCTURE 2021 State of Cloud HPC Report

For best workload performance, both architecture and

cloud provider decisions matter
Rescale Performance Index range by workload class

1.6

1.4 Most leading cloud

providers have a range of
1.2
architectures best suited
1 for different workloads.
0.8
Range of per-core
While the performance
0.6 performance for this of each CSP architecture
provider’s coretypes
0.4 against FEA workloads portfolio varies, it’s equally
important to find the best
0.2 architecture within the
0 portfolio to get the best
Computational Fluid Dynamics (CFD) Finite Element Analysis (FEA) Molecular Dynamics (MD) performance.
CSP A CSP B CSP C CSP D

About RPI: The Rescale Performance Index (RPI) is a hardware scoring system that analyzes leading benchmarks on common
software to enable performance comparisons between coretypes, workloads, and core scale
10
HARDWARE & INFRASTRUCTURE 2021 State of Cloud HPC Report

Half of organizations pursue multi-cloud strategies

Multi-cloud posture of organizations’ HPC practices

Fifty percent of organizations take

80%
adva
50% multi-cloud adoption
(up from 35% in 2018) dvantage of
60% 65% multi-cloud
Percentage of Organizations

50%
40%
43%

33%
20%

2% 1%
6% 0%
0%
1 CSP 2 CSPs 3 CSPs 4+ CSPs

2018 2020

Source: Rescale Cloud HPC Platform Analytics 11

2
2021 State of Cloud HPC Report

Software & Workloads

Key Findings Spotlight: Arrival

• Enterprises are running several applications across

ISV vendors

• Organizations often standardize on specific

software versions

• AI / ML workloads are increasingly common in HPC “Arrival works with Rescale to manage a broad software
portfolio and ensure engineers have access to the right
applications, the latest versions, and the most performant
hardware across each simulation requirement.”

- Nathan Baker, Head of Rapid Engineering Design -

Central Analysis Team, Arrival

12
SOFTWARE & WORKLOADS 2021 State of Cloud HPC Report

Leading organizations utilize a breadth of software

and versions
Number of software run by organizations Number of software versions run by organizations

50% 50%

49%
40% 40% 45%
Percent of Organizations

Percent of Organizations
30% 30%

20% 25% 20%

17% 16%
17%
10% 10% 12%
9%
0%
9% 0% 2%
1-2 3-4 5-9 10+ 1 only 2-5 6-10 11-20 21-50 51-100
Software Used Software Versions Used

Most organizations run three or more software packages, in While many run any number of software versions, most
many cases to support multidisciplinary simulation efforts. organizations standardize on a small number of versions.

Source: Rescale Cloud HPC Platform Analytics 13

SOFTWARE & WORKLOADS 2021 State of Cloud HPC Report

Leading R&D software being adopted in the cloud

Top 5 Used Software Applications, Sorted Alphabetically, Excludes In-House SoftwareExcludes In-House Software

Aerospace Automotive Energy Higher Education

Ansys Fluent Ansys Fluent Ansys Fluent Bring Your Own Container
CFD++ CONVERGE CFD LS-DYNA LAMMPS
Hexagon / MSC Software Nastran LS-DYNA MathWorks MATLAB Cresset Flare
Open Source CFD Solvers STAR-CCM+ Orcina OrcaFlex NWChem
Simcenter STAR-CCM+ Virtual Performance Solution Simcenter STAR-CCM+ VASP

Adoption of commercial Automakers’ software OrcaFlex gained popularity Use of containers is

codes like Fluent and preferences vary by region. as an industry standard for championed by higher
STAR-CCM+ surges in the VPS and LS-DYNA lead in use in a wide range of oil education for collaboration
R&D of next generation air finite element analyses. & gas analyses. and mobility across
and space vehicles. compute environments.

Source: Rescale Cloud HPC Platform Analytics, Note: Excludes In-House Software 14
SOFTWARE & WORKLOADS 2021 State of Cloud HPC Report

Leading R&D software being adopted in the cloud

Top 5 Used Software Applications, Sorted Alphabetically, Excludes In-House SoftwareExcludes In-House Software

Heavy Industry Life Sciences Machine Learning Semicononics

Ansys Fluent Gaussian Bring Your Own Container Cadence Virtuoso

CONVERGE CFD GROMACS Keras Hexagon / MSC Software Cradle

Exa PowerFLOW LAMMPS Pytorch Mentor Graphics Calibre

Simcenter STAR-CCM+ M-Star CFD R Cadence Innovus

SIMULIA Abaqus NAMD Tensorflow Synopsys ICV

Exa PowerFLOW and Growth seen in EDA

CONVERGE CFD see workloads shifting to the
significant growth for cloud as they run and scale
industrial combustion up efficiently on the latest
workloads. cloud hardware.

Source: Rescale Cloud HPC Platform Analytics, Note: Excludes In-House Software 15
3
2021 State of Cloud HPC Report

Performance Management
Key Findings Spotlight: Denso

• Each cloud provider brings its unique strengths to

high performance computing

• Finding the most performant architecture for a

workload is often a tradeoff between maturity,
performance, and value
“Denso runs cloud-based high-performance computing at
scale, and we are always investigating ways to optimize
• Full-stack optimization is the best approach to balance performance and cost savings. Rescale’s data-driven control
plane approach helps us complete our simulation jobs faster
cost and time-to-solve objectives than ever, while remaining very cost-effective.”

- Nate Moss, Senior CAD Analyst, Denso

16
PERFORMANCE MANAGEMENT 2021 State of Cloud HPC Report

Rescale’s data-driven approach to HPC management

Getting the most out of cloud HPC means starting from the software and
using the best architecture to run each workload. This requires a data-
Performance driven approach to assessing a large number of workloads against an ever-
Hardware performance by workload growing number of hardware architectures and various price points for each
and common job configurations architecture.

To accomplish this, Rescale has developed a structured approach

to assess performance, maturity, and value for architectures under
RPI management on its platform. Components include:

• The Rescale Performance Index (RPI) - a composite score of HW

performance that incorporates workloads, common benchmarks, and
various scales to facilitate performance comparisons by coretype
• The Rescale Maturity Index (RMI) - a hardware rating based on an
architecture’s track record, capacity, and volume of core hours under
RPI / operation
RMI Cost
• Value - a combination of RPI divided by price of the infrastructure,
which can vary across geographies and regions

Maturity & Scale Value

With this information, CIOs and IT / HPC managers can identify the full
Capacity, resilience & recent Performance per unit cost
economics and service level impact of their decisions, empowering them to
incident history
define policies and focus on accelerating time to answer, minimizing full-stack
SW/HW costs, or simply getting the most out of their existing budgets.

17
PERFORMANCE MANAGEMENT 2021 State of Cloud HPC Report

Leading provider with greatest breadth

of architectures
A workload where an AWS coretype provides competitive per-core
performance and value, with unmatched capacity Overview

AWS has an estimated ~50% share of the overall cloud computing market
Performance
and recently made $13B in cloud infrastructure investments in 2020 to fulfill
future growth and demand such as plans for a new region in Osaka. AWS
focus on HPC has been to deliver faster results with flexible configurations
and cost-effective operations.
CSP 1
Recent Developments
CSP 2
CSP 3 • Elastic Fabric Adapter (EFA), a unique OS bypass networking
mechanism that provides a low-latency, low-jitter channel for
AWS inter-machine communications
• Fully managed Lustre offering providing high performance and
scalable storage
Maturity & Scale Value
• Significant choice in architecture options (from high memory, high
storage, to compute-optimized) due maturity of HPC offerings
• Custom silicon with their 64-bit ARM Neoverse Graviton processors
that can deliver up to 40% better price performance
Source: Rescale Cloud HPC Platform Analytics, Using AWS list prices, Workload based on Caravan
benchmark on LS-DYNA
18
PERFORMANCE MANAGEMENT 2021 State of Cloud HPC Report

Fast growing provider focused on

the enterprise
A workload where an Azure coretype delivers best per-core and
per-cost performance while offering competitive capacity Overview

Azure has an estimated ~26% share of the overall cloud computing market
Performance and recently made $20B in cloud infrastructure investments in 2020 to
fulfill future growth. Recently Azure has announced investments in new
locations including Seoul, Delhi, Doha, Melbourne, and Toronto. Azure’s
focus on HPC has been to deliver optimized performance, cost control,
agile production-class platform, and coordinated partner go-to-market
CSP 1 strategies.
CSP 2
Recent Developments
CSP 3
Azure • Announced HBv2 VMs generally available in 2020, the first in the
public cloud with 200 Gbps HDR InfiniBand from Mellanox
• Released creative solutions for the HPC market such as convertible
Maturity & Scale Value reserved instances model which allows customers to use the latest
and greatest architectures while benefiting from reserve instance
pricing
• Hardware offerings include a wide variety of hardware offerings to
target your specific needs, including CPUs, GPUs, and FPGAs
Source: Rescale Cloud HPC Platform Analytics, Using Azure list prices, Workload based on Oil Rig
benchmark on Ansys Fluent
19
PERFORMANCE MANAGEMENT 2021 State of Cloud HPC Report

Helping bare-metal make a come-

back in the cloud
A workload where an OCI coretype delivers highest per-core and
cost performance
Overview
Performance
Oracle Cloud Infrastructure recently made $1.5B in cloud infrastructure
investments in 2020. OCI HPC platform includes bare metal compute
instances, low latency cluster networks for RDMA and advanced
technologies such as high-performance storage solutions and filesystems.
CSP 1 OCI brings an on-prem approach to HW with cloud-like procurement with
bare metal platform.
CSP 2
CSP 3 Recent Developments

Oracle • Flexible architecture enables multiple configurations (E3.Flex)

• Bare-metal approach to HPC for lowest latency with high throughput
interconnect
Maturity & Scale Value

Source: Rescale Cloud HPC Platform Analytics, Using Oracle list prices, Workload based on
Molecular Dynamics LJ Liquid 2M on LAMMPS
20
PERFORMANCE MANAGEMENT 2021 State of Cloud HPC Report

Focused on machine learning

performance
Google demonstrates leadership in machine learning workloads Overview

Google Cloud Platform (GCP) estimates serving ~10% share of the overall cloud
computing market and recently made $26B in cloud infrastructure investments
Google Sets Six Large Scale Training Performance Records in MLPerf v0.7
in 2020 to fulfill growth. New regions planned include Israel, Mexico, and
Higher is better; results are normalized to fastest non-Google submission
Spain. Google Cloud’s flexible and scalable offerings help accelerate
3 time to completion, so organizations can convert ideas into discoveries
and inspirations into products. GCP’s focus on HPC market involves taking
advantage of open source, going to market with partners, and introducing
2 advanced technologies especially in the machine learning area with GPUs &
TPUs investments.
1
Recent Developments

0 • First to release Nvidia’s Ampere A100 Tensor Core GPU in their

DLRM Transformer BERT SSD ResNet-50 Mask R-CNN A2 family, offering up to 20x the compute performance over prior
Google Other generations
• Custom machine configurations ensure orgs can get the hardware
specs engineers need, no more, no less
• Compute optimized (C2) instance offering and compact placement
policies help ensure lower latency topologies
Source: Google, “Google breaks AI performance records in MLPerf with world’s fastest training
supercomputer” June 29, 2020
21
PERFORMANCE MANAGEMENT 2021 State of Cloud HPC Report

The economic impact of full-stack optimization

Full-Stack Economics Analysis for a Crash Testing Use Case (FEA) Across Hardware Options
Baseline Full-Stack Cost Alternate Hardware Scenario Costs
$Million
$12.00 $10.87
$9.73 $9.89
$10.00 $8.60 $8.53
$4.75
$5.05 $5.08 $7.60
$8.00
SW
$5.31 $4.81
$6.00 $4.66
HW $6.12
$4.00 $4.81
$4.68
$2.00 $3.29 $3.72
$2.94
$0.00
Current Alternate HW 1: Alternate HW 2: Alternate HW 3: Alternate HW 4: Alternate HW 5:
Hardware Cloud Data Center Cloud Data Center Cloud
Workload Performance
(Time to Completion)
7 days 4 days 6 days 7 days 8 days 10 days

Fastest Reduced Cost, Lowest Cost

Maintained Performance

Full-Stack management can Counterintuitively, more expensive Optimization can be periodic

reduce spend by while maintaining hardware can sometimes lead to (e.g., annual), or automated
performance, or accelerate time-to- lower software spend and continuously as pricing and new
answer at the same or lower costs. overall lower costs. architectures become available.

Source: Rescale Cloud HPC Platform Analytics 22

4
2021 State of Cloud HPC Report

Business Management
Key Findings Spotlight: Aerion

• Enterprises apply financial controls to mitigate

runaway cloud costs

• Organizations across all segments often use SW &

HW policies to apply best practices

• Security and compliance are key concerns for R&D-

“Rescale gives us compliance, control, and economic
visibility with the best of the cloud, allowing us to work
quickly, efficiently and in parallel with each other” said Bissell
driven organizations Smith, CIO & EVP at Aerion Supersonic. “Cloud computing
is a fundamental part of Aerion’s long-term strategy. If
compliance isn’t met, the consequences for the company,
employees, and even the country can be severe, so we take
it very seriously.”

- Bissell Smith, CIO & EVP at Aerion Supersonic.

23
BUSINESS MANAGEMENT 2021 State of Cloud HPC Report

Enterprises apply financial controls to mitigate runaway

cloud costs
Organizations using financial controls and number of budgets set

100% 20.0 As cloud HPC adoption

increases across the
organization, financial controls
Percent of Organizations

75% 15.0
prevent costly overruns.

50% 10.0
Separate and specific budgets
at a BU, team, project and user
25% 5.0 level help organizations adopt an
efficient cloud OPEX model.
0% 0.0
SMB Commercial Enterprise Organizations with financial
% Using Cost Controls # of Budgets Set
transparency and control
can more accurately forecast
resource needs and assess ROI.

Source: Rescale Cloud HPC Platform Analytics 24

BUSINESS MANAGEMENT 2021 State of Cloud HPC Report

Policy-based software and hardware controls

commonly adopted in each segment
Percent of organizations using hardware and software policy controls

80%
Organizations seek to
standardize optimal hardware
and software combinations for
60%
64% each specific workload.
52%
40%
45%
50% Larger organizations with more
40% HPC groups show increased
35%
20%
usage of hardware and software
policies.

0%
SMB Commercial Enterprise HPC managers are identifying new,
more performant configurations
Using Software Policies Using Hardware Policies
and setting policies to standardize
Company Size (SMB = 1 - 2 Groups, Commercial = 3 - 5 Groups, Enterprise = 6+ Groups) best practices.

Source: Rescale Cloud HPC Platform Analytics 25

BUSINESS MANAGEMENT 2021 State of Cloud HPC Report

Security and compliance are key enterprise concerns

Top Functional Interests Top Certifications of Interest

Authentication Identity & access

management

Data security & Data governance

encryption

Commercial and Enterprise organizations require Organizations with collaboration and multiple shared
additional security measures beyond the standard workspaces implement a high level of security to
cloud provider shared responsibility security model. protect intellectual property and user data.
26
5
2021 State of Cloud HPC Report

Practitioner Sentiment
Key Findings Spotlight: UWE Bristol

• Managing data is a top challenge for both HPC users

and managers

• HPC / IT leaders are looking to deliver more holistic

R&D services including workflow automation

• HPC managers and users both predict aggressive Rescale modernized our School of Engineering’s HPC
capabilities with a control plane for hybrid and multi-cloud
shift to cloud in the next five years infrastructure to accelerate our simulations on one easy-to-
use platform.”

- Dr. Pritesh Narayan, Aerospace, Aviation and Management

Cluster Leader, University of the West of England

27
PRACTITIONER SENTIMENT 2021 State of Cloud HPC Report

Data management and computing capacity are top

challenges
Users directly feel the pain of
What are the top challenges to maintaining an optimal HPC environment?
data management and waiting
30% for jobs to complete, and HPC
30%
managers agree.
25% 25%
Percent of Respondents

20% 22% Software license management

complexity can cause
14% 14% 14% engineering delays, imprecise
13% 13% 13%
10%
11% investments, and suboptimal
8% utilization.
0%
Moving data Long queuing time / Shortage of Insufficient budget Slow simulation Limited Distributed teams face new
around - managing Limited availablity software licenses time collaboration and
input / output files data sharing collaboration challenges such
Challenges
as securely and efficiently
sharing data in the cloud.
Users Managers

Source: 2020 HPC Practitioner Survey 28

PRACTITIONER SENTIMENT 2021 State of Cloud HPC Report

Capacity, visibility, and automation are key

investment areas
What are the highest priority HPC initiatives in the next 12 months? Select 3 (Managers)

50%

The need to modernize HPC

44%
40%
42% infrastructure and deliver more
Percent of Respondents

39% 39%
36% holistic R&D solutions such as
30% 33%
workflow automation is tempered
20%
25% by the HPC talent gap.

10%
To do more with less, many IT/HPC
leaders are looking to cloud HPC
0%
Increase
Capacity
Find More
Talent
Automate
Workflows
Upgrade
Infrastructure
Improve
Reporting
Migrate to the
Cloud
Reduce Overall
cost
management and automation.
Investment Category

Source: 2020 HPC Practitioner Survey 29

PRACTITIONER SENTIMENT 2021 State of Cloud HPC Report

Cloud expected to dominate in a hybrid future

Where will HPC workloads be run in the next 5 years? (Managers and users combined)

50%

40%
42%
Percent of Respondents

While HPC practitioners

30%
have been slower to adopt
31%
cloud compared to broader
20%
enterprise IT, most now predict
19%
17% a cloud-centric future.
10%

0% 0%
Almost Entirely Mostly Equal Parts Mostly in the Almost Entirely in
On-Premises On-Premises On-Premises Cloud the Cloud

Predictions

Source: 2020 HPC Practitioner Survey 30

PRACTITIONER SENTIMENT 2021 State of Cloud HPC Report

Organizations remain cautiously optimistic on cloud

What is the biggest concern about migrating workloads to the cloud? (Managers)

50%

40% 44% 44% Cost, complexity, and security

are top concerns when moving
Percent of Respondents

39%
30%
workloads to cloud.

22% Organizations consider ways

20%
22%
20%
to orchestrate, automate, and
optimize cloud HPC to mitigate
10% 14% their key concerns.
11%

0%
Cost Complexity Security No Concern Maintenance / Usability
Management
Cloud Concerns

Source: 2020 HPC Practitioner Survey 31

 2021 State of Cloud HPC Report

Intelligent Computing for Digital R&D

About Rescale
Rescale helps organizations accelerate science and engineering Learn how you can modernize your high performance computing:
breakthroughs by eliminating computing complexity. From
supersonic jets to personalized medicine, industry leaders bring new
product innovations to market with unprecedented agility, speed,
and efficiency with the Rescale Platform - an intelligent full-stack Get in touch Start your View
automation HPC solution for digital R&D in the cloud. IT leaders use with our team free trial additional
Rescale to deliver high-performance computing-as-a-service to their of experts today resources
organization by harnessing the power of automation on a hybrid cloud
control plane with security, architecture, and financial controls.

www.rescale.com
32