NVIDIA-powered Data Science Workstations could change everything

Source: NVIDIA. A data science workstation.

NVIDIA has teamed with the world’s leading OEMs and system builders to deliver workstations designed to help data scientists, analysts and engineers become more productive.

Data science problems such as predicting outcomes for project management, monitoring the health of equipment, and flight simulation.involve data on a massive scale and require large-scale processing capabilities. The new NVIDIA-powered data science workstations are purpose-built for data analytics, machine learning and deep learning, and provide the extreme computational power and tools required to prepare, process and analyse massive amounts of data. This will allow users to make better business predictions, faster.

NVIDIA-powered workstations for data science are based on a reference architecture consisting of dual high-end NVIDIA Quadro RTX GPUs and NVIDIA CUDA-X AI accelerated data science software, such as RAPIDS, TensorFlow, PyTorch and Caffe. CUDA-X AI is a collection of libraries that enable modern computing applications to benefit from NVIDIA’s GPU-accelerated computing platform.

Data science processing is faster with GPUs.

NVIDIA CEO Jensen Huang said, “Data science is one of the fastest growing fields of computer science and impacts every industry. Enterprises are eager to unlock the value of their business data using machine learning and are hiring at an unprecedented rate data scientists who require powerful workstations architected specifically for their needs.

“With our partners, we are introducing NVIDIA-powered data science workstations, made possible by our new Turing Tensor Core GPUs and CUDA-X AI acceleration libraries, that allow data scientists to develop predictive models that can revolutionise their business.”

Features and benefits include:

• Dual, high-end Quadro RTX GPUs

Powered by the latest NVIDIA Turing GPU architecture and designed for enterprise deployment, dual Quadro RTX 8000 and 6000 GPUs deliver up to 260 teraflops of compute performance and 96 GB of memory using NVIDIA NVLink interconnect technology.

• Data science software stack 

Built on Linux and Docker containers, the software features:

• NVIDIA CUDA-X AI — A collection of NVIDIA's GPU acceleration libraries to accelerate deep learning, machine learning (ML) and data analysis. CUDA-X AI includes cuDNN for accelerating deep learning primitives, cuML for accelerating machine learning algorithms, TensorRT for optimising trained models for inference and over 15 other libraries.
• NVIDIA RAPIDS — A set of GPU-accelerated libraries analytics for data preparation, traditional machine learning and graph analytics.
• Anaconda Distribution — With Anaconda, NVIDIA is providing Anaconda Distribution, which allows data scientists to perform Python/R, data science, AI and machine learning. \

• Enterprise ready 

Tested and optimised in conjunction with workstation manufacturers to meet the needs of mission-critical enterprise deployments.

• Optional software support 

Offers peace of mind with NVIDIA-developed software and containers, including deep learning and machine learning frameworks.

NVIDIA-powered systems for data scientists are available from global workstation providers such as Dell, HP and Lenovo and regional system builders, including AMAX, APY, Azken Muga, BOXX, CADNetwork, Carri, Colfax, Delta, EXXACT, Microway, Scan, Sysgen and Thinkmate.

“The value of understanding data in running a business is clear, yet there is a current lack of tools,technology and education in data analytics and data exploration. HP Z Workstations, powered by Quadro RTX and ​RAPIDS, p​rovide the performance and power to help our customers benefit from bigdata – all in an easily deployable solution,” said Bruce Blaho, VP, HP Fellow and Chief Technologist for Z by HP at HP Inc.

A HP data science workstation in action.

“Powered by NVIDIA Quadro RTX professional GPUs and offering an NVIDIA RAPIDS-ready platform, our Lenovo ThinkStation ​​AI Workstation​ offers users who face challenging machine and deep learning models the ability to dramatically transform the efficiency of their workflow,” added Rob Herman,General Manager of Workstations and Client AI at Lenovo.

“Artificial intelligence (AI) and ML technologies are game-changers for businesses embarking on digital transformation efforts. The systems our customers rely on are becoming smarter. This, along with increased computing power and capabilities are helping all of us make better informed decisions that are leading to tangible business outcomes.

"For organisations of all sizes, there’s a huge opportunity to speed up processes and deliver tailored services to increase customer loyalty, engagement and satisfaction. Dell is in a unique position to offer edge to core to cloud solutions and expertise to customers looking to implement AI and ML into their workloads. Dell Precision workstations, in combination with NVIDIA’s reference architecture for data science, will help to streamline data into useful and actionable information while updating IT infrastructure with technologythat best suits a customer’s needs,” said Tom Tobul, VP, Specialty Commercial Client Solutions at Dell.

In a demonstration of HP's Z series of workstations for data science at NVIDIA's GPU Technology Conference (GTC) a booth representative said that it would have been impossible in the past to present visualisations from the demonstrated dataset because of its size and scale.

"It would just have crashed," the representative explained.

"We have a diverse, multidisciplinary environment and are looking to couple data science and analytics to a wider range of our technical practices throughout our business. The NVIDIA-powered Data Science Workstation promises to ease the transition and democratise the application of data science.

"We find it extremely well-suited to experimentation, exploration, solution discovery and early prototyping work. Its combination of well-designed software and highly performant hardware provides 20x and higher speedups in our analytics work and our team found its ease-of-use liberating,” said Steve Walker, Associate Director of Advanced Digital Engineering at Arup.

“The NVIDIA-powered data science workstation enables our data scientists to run end-to-end data processing pipelines on large datasets faster than ever,” said Mike Koelemay, Chief Data Scientist at Lockheed Martin Rotary & Mission Systems.

“Leveraging RAPIDS to push more of the data processing pipeline to the GPU reduces model development time, which leads to faster deployment and business insights.”

“Our initial look at the NVIDIA-powered Lenovo AI workstation showed significant performance gains. Data scientists will appreciate being able to move more quickly through the analytics life cycle, which will allow them to address and support more analytics needs to transform business processes,” said Gavin Day, Senior Vice President for Technology at SAS.

“Delivering near real-time data science is a game-changer when it comes to making sense of our network data. Before, we had access to terabytes of data daily with no efficient way to gain insights out of it. Now, every time we look at the data, we see something new that we can take immediate advantage of. This is made possible by NVIDIA-powered Data Science Workstations at our desks, said Jared Ritter, Director of Wireless Engineering at Charter Communications.

"The combination of RAPIDS and software from Datalogue and OmniSci completely changes the way we collect, process, visualise and understand data. We are able to build models to predict high-surge Wi-Fi usage, offload access points swiftly and streamline operations to save millions of dollars.” NVIDIA's RAPIDS software libraries, built on CUDA-X AI, enable end-to-end GPU-delivered data science and analytics.

Ritter shared in a separate panel at GTC that before adopting NVIDIA AI technology, it had been impossible to get insights in real time from corporate data. The company is the result of mergers of different companies, leading to a pool of very disparate data sets derived from different network topologies. And while the company also has domain experts, human experts would take a week to make sense of data whereas decision makers cannot wait a week, he said.

“Having bandwidth to ingest all that data, visualising all that information.. that was amazing,” he commented.

Charter could use its system to show the impact of a wildfire, where access points are up or down, and where people might access Wi-Fi for free, Ritter said.

“This is capability that we've never had before – (to get an) impact assessment that quickly, like in real time. It could mean the difference to people who need (a) service, especially in a disaster.”

The democratisation of data science enabled by the workstations was a key theme during the panel. "Augment human decisions – that's what these workstations will do," observed Sertac Karaman, Associate Professor of Aeronautics and Astronautics at the Massachusetts Institute of Technology.

Day, who also spoke on the panel, noted that the workstations move data science work out of a data centre and onto office desks. “It's accessibility, lowering the barrier of entry. This is where it'll take off from a market perspective,” he said.

Koelemay, who was part of the panel as well, agreed. “It's a force multiplier. It can do the work of a bunch of engineers doing manual ETL, and you can do it in real time.” ETL stands for extract-transform-load, which are data processing tasks required for analytics.

Panelists also touched on solutions for the dearth of data science skills. Koelemay noted that data scientists are described as unicorns. “They exist but are hard to find,” he said.

That said, data scientists do not work in a vacuum. A data science team comprises specialists with different skillsets, such as software engineers, data engineers, people who understand schemas, storage mechanisms, and systems, as well as generalists, who coordinate workflow, he said.

Rather than train data scientists in domain expertise, Lockheed Martin has data-science upskilling programmes and internal training for domain experts. “We're seeing a lot of data science programmes starting up. Our best hires are (people who major in) dual computer science and mathematics,” he said.

“There's still a massive skills gap there,” Day said. “Having this skill, building it up internally is something that we see our customers are spending a lot of time on. If the vision is realised here with data science workstations, all of a sudden there's something that we can move forward on.”

Hashtag: #GTC19

*NVIDIA sponsored transport and accommodation to GTC.