Data is a strategic tool for life sciences organizations looking to accelerate their research with digital insights to create faster outcomes. Existing IT and data storage systems were simply not designed to handle the very high-speed and massive-scale requirements of AI and analytics that these organizations need to succeed. Success in today’s AI-driven world depends on access to a new high-performance, data-centric IT approach. Organizations are looking at a variety of deployment models to address these requirements.
Cambridge-1, the UK’s most powerful supercomputer, enables top scientists and healthcare researchers to use the combination of AI and simulation to speed the digital biology revolution and bolster the country’s world-leading life sciences industry. Cambridge-1 is the fastest AI supercomputer in the UK, with 80 NVIDIA DGX A100 systems backed by more than 2PB of DDN A3I storage for AI workflows. Cambridge-1 is the first of NVIDIA’s supercomputers to be made available to external researchers.
An NVIDIA DGX SuperPOD™, Cambridge-1 is solving the most pressing medical challenges and accelerating health research spanning medical imaging, genomics, and drug discovery. With its founding partners — AstraZeneca, GlaxoSmithKline, Guy’s and St Thomas’ NHS Foundation Trust, King’s College London, and Oxford Nanopore — and other UK organizations, Cambridge-1 is dedicated to advancing UK healthcare research through digital biology, unlocking a deeper understanding of disease and breakthroughs through medicine.
Watch as Marc Hamilton, vice president of solutions architecture and engineering at NVIDIA, explains more about Cambridge-1, the importance of a proper infrastructure, and why NVIDIA chose DDN for the AI storage of this supercomputer.
The Modern Instrument for Drug Discovery
Organizations utilizing traditional enterprise storage systems when trying to implement AI and analytics workflows are experiencing difficulties in getting these programs into production. Current systems introduce bottlenecks in the workflow and lockups that choke performance and don’t allow large multinode systems, like the NVIDIA DGX SuperPOD, to address large datasets at maximum utilization. With AI and high-performance computing becoming essential tools of modern healthcare and life science discovery, there is no time to waste on slow-running applications, underutilized GPUs, or difficult-to-manage systems. To speed up research by orders of magnitude utilizing analytics and AI, institutions need to store data cost effectively and process data quickly.
Hamilton reminds us that storage is often an overlooked component to AI deployments and provides top requirements for AI storage:
- Performance
- Capacity
- Parallel Data Movement
- Ease of use
Watch this video for more detail on why each of these are so important to the success of AI projects.
“There are many important considerations when designing this sort of infrastructure to power the world’s most powerful AI systems. Storage is one that’s often overlooked.”
NVIDIA Selects DDN’s AI Solutions to Turn Data into Knowledge
DDN solutions allow life sciences leaders to advance their research faster and at a lower cost with a centralized, simple system that can manage the complete AI data life cycle.
Hamilton elaborates, “When that researcher has that idea and they’re ready to go, the storage needs to be available. They can’t be going and looking for a backup copy, worrying about moving storage from different devices to other devices. And that’s what the DGX SuperPOD with DDN storage accomplishes. It’s a very simple architecture. With GPU Direct Storage, we’re able to go through and move the data directly from the DDN AI400X over the network and move it directly from the InfiniBand interface into the GPUs’ memory. This provides significant time savings and increases the performance of the overall system.”
“I never hesitate to recommend DDN. I know that if DDN can meet the demands of the DGX SuperPOD, it can meet any of our customers’ requirements.”
DDN’s long-standing relationship with NVIDIA is demonstrated through NVIDIA’s choice to use DDN’s scalable technology for its supercomputers, knowing the infrastructure will deliver for customers. The solution addresses the requirements of production AI environments by removing the complexity and uncertainty with a proven combination of industry-leading intelligent storage and GPU technology. By simplifying the selection, configuration, purchase, and deployment of AI infrastructure with reference architectures, DDN, using NVIDIA technology, can immediately boost AI application performance while eliminating the management complexity and poor performance that can stall AI programs and research time.
“The real differentiator is DDN’s use of a parallel file system – that’s why DDN storage was the first to be qualified with DGX SuperPOD and with EXAScaler, DDN provides the same class enterprise-grade service and support that you would expect with any commercial storage offering,” noted Hamilton.
This means customers’ applications will run quickly, ensuring the infrastructure is fully utilized and organizations avoid implementation hurdles and IT overhead. The brilliant work starts here, with production-proven reference architectures that make the rollout of AI initiatives easier – and faster – so operational AI services can be implemented in weeks, rather than months or years.
Watch this video and take a deeper look inside Cambridge-1 as Marc Hamilton shares its importance and why NVIDIA chose DDN to support the fastest AI supercomputer in the UK:
“DDN is the de facto name for AI storage in high-performance environments.”
Accelerate your AI transformation at NVIDIA GTC, a virtual conference running March 21-24, and learn how organizations like NVIDIA, and many others, have successfully implemented AI with DDN. Speak to a storage specialist today and turn data into knowledge.
