Solution Brief

Dell EMC Ready Solutions for Virtual Desktop

Infrastructure
Utilizing hyper-converged systems for GPU resource sharing for VDI and AI workloads in healthcare (Full paper here)

Benefits of optimizing workload consolidation

Challenges to running parallel mixed
workloads There is no doubt that many industries have benefited
significantly from embracing virtual desktop infrastructure (VDI)
given all the benefits to data control, security, mobility and more.
GPU-accelerated VDI has been heavily adopted specifically in
Clinician Using PACS AI Training and areas with large design or image processing needs. Healthcare
Viewer Inferencing is one such industry due to the need for high performance image
processing in departments such as radiology, cardiology, and
digital pathology.

VDI Workstation AI Workstation At the same time, many companies have begun to benefit a
great deal from GPU-accelerate artificial intelligence initiatives.
These programs, while clearly beneficial to the long-term
outcome for the company, are often difficult to commit to and to
deploy at scale. IT administrators face barriers such as:
Healthcare has increasingly begun utilizing
• Insufficient scale to fully utilize GPU
machine learning systems to improve • Difficulty sizing AI deployment to match user input
countless aspects of the value chain, from • Systems complexity in transferring large image files
improved image categorization and storage • Cost barriers in deploying new AI environments
to AI assisted cancer screenings.
These challenges are not uncommon for new private AI
Traditional large-scale means of deploying initiatives that struggle to find the right balance of utilization and
AI pose challenges for many in the scale. It can be difficult to justify sizing out large GPU farms for
healthcare industry: AI workloads until there are known users feeding data into that
platform, and it can be challenging to uncover how users will
• Insufficient scale to deploy dedicated feed data into the platform until they see the benefits and have it
VDI and AI environments made available to them. This ‘chicken-egg’ problem is common
in most AI projects and the typical solutions of outsourcing AI
• Irregular workloads lead to inability to compute resources to the public cloud or sizing very small in a
optimize GPU utilization proof of concept like manner, both present challenges to in the
form of data security and compliance issues in the public cloud,
• Latency challenges in moving large data and in cost structures and GPU utilization on premises. The key
files over network in relevant time to balancing these problems lie in tying AI workloads and the VDI
frames workloads to a common platform. By doing so, healthcare
institutions can increase the GPU utilization of existing cards
• Cost barriers in embracing AI within running in their VDI environments, improve system performance
compliance constraints of healthcare and responsiveness, and easily right size AI workloads by
data matching them to the platforms that need them.
But while most IT owners schedule these workloads at different
times of the day or week, we believe there is a strong case for an
alternative strategy.
 Beyond “VDI by day, compute by night”
Of course, using existing platforms for AI and ML initiatives is nothing new. Due to the nature of VDI, IT
administrators have often used common compute platforms for VDI implementations during working hours and
repurposing that infrastructure for training AI models or other batch data processing workloads. This schema is
not simple to implement, but can be helpful in maximizing utilization of existing infrastructure that would
otherwise be sitting idle. However, this relationship presumes the VDI environment itself is capable of
maintaining high utilization during the day and the AI workloads doing the same at night. This assumption may
not hold up in healthcare, where image processing requires significant amount of GPU’s for short periods of time.
That’s why we’ve architected and tested a shared VDI and AI infrastructure for healthcare where VDI and AI are
run as dual workloads simultaneously on the same platform. By running these workloads together, you can
achieve improved mobility and consistency of performance for your users, while utilizing the power of artificial
intelligence, all on the same common platform. Our guidance helps you to size and deploy this setup with
confidence, knowing our engineers have tested these configurations with the most popular AI frameworks and
VDI best practices.
By taking advantage of our free best practices guide, you can right size your AI deployments for your operations.
Our guidance helps you plan and manage this transition yourself through our best practices guide or we can help
manage this for you with our turnkey deployment services, simplified lifecycle management and single vendor
support experience from Dell Technologies. We are committed to helping customers run their workloads how
they work best for you and this guidance will help you make those choices in your business.
The Dell Technologies Difference
Dell Technologies has a deep expertise in both VDI and AI systems and can help you deliver on all your IT
initiatives. With strong partnerships with NVIDIA, Intel, and many leading AI frameworks to ensure our customers
can get the best performance and support for all their GPU deployments. We believe that by implementing a
shared common platform for VDI and AI workloads, organizations can increase their GPU utilization, lower their
TCO, improve the mobilities of their employees, and experience high system performance for both VDI graphics
and AI training and inferencing, thereby increasing the overall efficiency of their day-to-day operating
environment.
Modern IT platforms from Dell Technologies help run the next generation of medical imaging applications by
delivering critical solutions that give healthcare organizations the ability to balance clinical requirements with the
ability to deliver unified image solutions. As healthcare organizations shift from disconnected PACS to
interoperable medical imaging solutions, organizations will gain the ability to analyze medical imagery in real time
with powerful AI systems, retain only the images allowed for compliance, and easily transfer and share data with
patients and partners, creating better outcomes for everyone.
To answer the challenge of unlocking the power of artificial intelligence in the healthcare industry and do so as
seamlessly as possible, Dell EMC Ready Solutions is here to provide a proven path that delivers outcomes that fit your
business.

Learn More about

Dell EMC Solutions for VDI Contact a Dell EMC Expert
and
Dell Technologies InfoHub

© 2020 Dell Inc. or its subsidiaries. All Rights Reserved. Dell, EMC and other trademarks are
trademarks of Dell Inc. or its subsidiaries. Other trademarks may be trademarks of their respective
owners. Reference Number: H17874.2