Advanced Computing in the Age of AI | Tuesday, July 23, 2024

Oracle Joins Shift to Cloud GPUs 

Public cloud vendors continue to add processing horsepower in the form of Nvidia GPU accelerators and the graphics processor vendor's Tesla datacenter GPUs.

Oracle announced this week it is adding Nvidia's Tesla GPUs to its public cloud a month after cloud rival Google announced plans to add Nvidia GPUs to its public cloud. Both are attempting to differentiate their cloud services by expanding enterprise access to GPU-based computing.

Starting now, Nvidia (NASDAQ: NVDA) said Oracle (NYSE: ORCL) would provide customers with cloud access to its Tesla P100 GPU accelerators. Don Johnson, Oracle's senior vice president for product development, also announced Tuesday (Oct. 3) during a company event that the cloud provider also would offer access to Nvidia's Tesla V100 datacenter GPUs based on the chip maker's Volta architecture.

The move follows Google's (NASDAQ: GOOGL) announcement in September that it would launch P100 GPUs in beta while K80 GPUs were made available on the Google Cloud Platform.

The Oracle announcement comes as the database specialist moves aggressively into the hyper-competitive public cloud market. "It’s important to have great apps, but great apps that work together, the most complete suite of [platform] services and the next generation of infrastructure as a service that all work together to complement each other,” Oracle CEO Mark Hurd said this week. "That’s what we’ve built out, and that’s what we now have."

Oracle becomes the latest public cloud vendor to embrace Nvidia's P100 GPUs for hardware acceleration in the cloud. Along with Google, cloud rivals IBM (NYSE: IBM) and Microsoft (NASDAQ: MSFT) both announced earlier this year they would incorporate the graphics processors into cloud offerings.

For its part, Google is stressing cloud GPUs as a way of accelerating workloads that utilize machine learning training and inference as well as geophysical data processing, genomics and other high-performance computing applications.

Oracle said its cloud offerings would deliver two P100 GPUs with Nvidia interconnects technology capable of delivering 21 teraflops of performance per instance. The capability targets the applications such as deep learning training and inferencing, accelerated analytics and high performance computing.

Each P100 cloud instance would deliver the performance of up to 25 conventional servers, reducing the cost of running for HPC and AI workloads, it added.

Oracle is playing catch-up in a cloud market that is expanding with the embrace of enterprise multi-cloud strategies. Hurd predicted in a keynote that application development would move wholesale to the cloud by 2025. Meanwhile, the Oracle CEO forecast that as much as 80 percent of production applications would reside in the cloud with a similar amount being spent on cloud services over the same period. Meanwhile, Hurd added, "all enterprise data will be stored in the cloud" by the middle of the next decade.

About the author: George Leopold

George Leopold has written about science and technology for more than 30 years, focusing on electronics and aerospace technology. He previously served as executive editor of Electronic Engineering Times. Leopold is the author of "Calculated Risk: The Supersonic Life and Times of Gus Grissom" (Purdue University Press, 2016).