NVIDIA debuts RAPIDS, its open-source GPU-acceleration platform for large-scale data analytics and machine learning



NVIDIA announced on Wednesday a GPU-acceleration platform for data science and machine learning, which enables even the largest companies to analyze massive amounts of data and make accurate business predictions at improved speed.

RAPIDS open-source software gives data scientists a giant performance boost as they address highly complex business challenges, such as predicting credit card fraud, forecasting retail inventory and understanding customer buying behavior. Reflecting the growing consensus about the GPU’s importance in data analytics, an array of companies is supporting RAPIDS — from pioneers in the open-source community, such as Databricks and Anaconda, to tech leaders like Hewlett Packard Enterprise, IBM and Oracle.

The RAPIDS suite of software libraries gives customers the ability to execute end-to-end data science and analytics pipelines entirely on GPUs. It relies on NVIDIA CUDA primitives for low-level compute optimization, but exposes that GPU parallelism and high-bandwidth memory speed through user-friendly Python interfaces.

RAPIDS also focuses on common data preparation tasks for analytics and data science. This includes a familiar DataFrame API that integrates with a variety of machine learning algorithms for end-to-end pipeline accelerations without paying typical serialization costs. RAPIDS also includes support for multi-node, multi-GPU deployments, enabling vastly accelerated processing and training on much larger dataset sizes.

RAPIDS offers a suite of open-source libraries for GPU-accelerated analytics, machine learning and, soon, data visualization. It has been developed over the past two years by NVIDIA engineers in close collaboration with key open-source contributors.

For the first time, it gives scientists the tools they need to run the entire data science pipeline on GPUs. Initial RAPIDS benchmarking, using the XGBoost machine learning algorithm for training on an NVIDIA DGX-2 system, shows 50 times speedups compared with CPU-only systems. This allows data scientists to reduce typical training times from days to hours, or from hours to minutes, depending on the size of their dataset.

Analysts estimate the server market for data science and machine learning at $20 billion annually, which — together with scientific analysis and deep learning — pushes up the value of the high performance computing market to approximately US$36 billion.

“Data analytics and machine learning are the largest segments of the high performance computing market that have not been accelerated — until now,” said Jensen Huang, founder and CEO of NVIDIA, who revealed RAPIDS in his keynote address at the GPU Technology Conference. “The world’s largest industries run algorithms written by machine learning on a sea of servers to sense complex patterns in their market and environment, and make fast, accurate predictions that directly impact their bottom line.

“Building on CUDA and its global ecosystem, and working closely with the open-source community, we have created the RAPIDS GPU-acceleration platform. It integrates seamlessly into the world’s most popular data science libraries and workflows to speed up machine learning. We are turbocharging machine learning like we have done with deep learning,” he said.

RAPIDS builds on open-source projects, including Apache Arrow, pandas and scikit-learn, by adding GPU acceleration to the Python data science toolchain. To bring additional machine learning libraries and capabilities to RAPIDS, NVIDIA is collaborating with such open-source ecosystem contributors as Anaconda, BlazingDB, Databricks, Quansight and scikit-learn, as well as Wes McKinney, head of Ursa Labs and creator of Apache Arrow and pandas, the fastest-growing Python data science library.

“RAPIDS, a GPU-accelerated data science platform, is a next-generation computational ecosystem powered by Apache Arrow,” McKinney said. “NVIDIA’s collaboration with Ursa Labs will accelerate the pace of innovation in the core Arrow libraries and help bring about major performance boosts in analytics and feature engineering workloads.”

To facilitate broad adoption, NVIDIA is integrating RAPIDS into Apache Spark, the open-source framework for analytics and data science.

“At Databricks, we are excited about RAPIDS’ potential to accelerate Apache Spark workloads,” said Matei Zaharia, co-founder and chief technologist of Databricks, and founder of Apache Spark. “We have multiple ongoing projects to integrate Spark better with native accelerators, including Apache Arrow support and GPU scheduling with Project Hydrogen. We believe that RAPIDS is an exciting new opportunity to scale our customers’ data science and AI workloads.”

Tech enterprises across a range of industries are early adopters of NVIDIA’s GPU-acceleration platform and RAPIDS.

“NVIDIA’s GPU-acceleration platform with RAPIDS software has immensely improved how we use data — enabling the most complex models to run at scale and deliver even more accurate forecasting,” said Jeremy King, executive vice president and chief technology officer at Walmart. “RAPIDS has its roots in deep collaboration between NVIDIA’s and Walmart’s engineers, and we plan to build on this relationship.”

Additionally, some of the world’s leading technology companies are supporting RAPIDS through new systems, data science platforms and software solutions:

“HPE is committed to advancing the way customers live and work. Artificial intelligence, analytics and machine learning technology can play a critical role in uncovering insights that can help customers achieve breakthrough results and improve the world we live in,” said Antonio Neri, CEO, Hewlett Packard Enterprise. “HPE is unique in the market in that we provide complete AI and data analytics solutions from strategic advisory to purpose-built GPU accelerator technology, operational support and a strong partner ecosystem to tailor the right solution for each customer. We are excited to partner with NVIDIA on RAPIDS to accelerate the application of data science and machine learning to help our customers drive faster and more insightful outcomes.”

“IBM has built the world’s leading platform for enterprise AI, regardless of deployment model,” said Arvind Krishna, senior vice president of Hybrid Cloud and director of IBM Research. “We look forward to extending our successful partnership with NVIDIA, leveraging RAPIDS to provide new machine learning tools for our clients.”

“The compute world today requires powerful processing to handle complex workloads like data science and analytics — it’s a job for NVIDIA GPUs. RAPIDS is accelerating the speed at which this processing and machine learning training can be done,” said Clay Magouyrk, senior vice president of Software Development, Oracle Cloud Infrastructure. “We are excited to support this new suite of open-source software natively on Oracle Cloud Infrastructure and look forward to working with NVIDIA to support RAPIDS across our platform, including the Oracle Data Science Cloud, to further accelerate our customers’ end to-end data science workflows. RAPIDS software runs seamlessly on the Oracle Cloud, allowing customers to support their HPC, AI and data science needs, all while taking advantage of the portfolio of GPU instances available on Oracle Cloud Infrastructure.”

MapR Technologies also announced support within the MapR Data Platform to accelerate data access and production deployments for data science through the RAPIDS open-source software.

MapR helps data scientists accelerate the access of required training data by focusing on easing the issues of on-boarding, cleansing, cataloging, and feeding data at high performance to GPUs and NVIDIA DGX systems. The MapR solution also manages the deployment and management of multiple models into production to speed business impact.

“The challenge for most data scientists is the data logistics to locate, prep and access the right data for training. In many cases, 90 percent of the time is spent data wrangling,” said Anil Gadre, EVP and chief product officer, MapR Technologies. “MapR complements RAPIDS with a data management and logistics fabric to accelerate the high-scale processing and access of disparate data across geographies. The same fabric also speeds the deployment of models into production and coordinates the continuous deployment and updating of multiple models to impact business in real-time at scale.”

Access to the RAPIDS open-source suite of libraries is immediately available online, where the code is being released under the Apache license. Containerized versions of RAPIDS will be available this week on the NVIDIA GPU Cloud container registry.

 

Leave a Reply

WWPI – Covering the best in IT since 1980