Nvidia ideas to purchase item storage vendor SwiftStack to assist its prospects accelerate their artificial intelligence, superior-performance computing, and info analytics workloads.
The Santa Clara-dependent GPU vendor will not sell SwiftStack software to new prospective clients but will use SwiftStack’s 1space as section of its inside artificial intelligence (AI) stack. It will also allow prospects to use the SwiftStack software as section of their AI stacks, according to Nvidia’s head of business computing, Manuvir Das.
SwiftStack and Nvidia disclosed the acquisition today. They did not reveal the invest in price tag, but they said it expects the offer to close with weeks.
Nvidia previously labored with SwiftStack
Nvidia labored with San Francisco-dependent SwiftStack for far more than 18 months on tackling the info issues affiliated with operating AI applications at a enormous scale. Nvidia observed 1space specifically practical. SwiftStack introduced 1space in 2018 to accelerate info accessibility throughout public and private clouds through a single item namespace.
“Only put, it is really a way of positioning the suitable info in the suitable position at the suitable time, so that when the GPU is busy, the info can be sent to it quickly,” said Das, who will oversee the integration of the SwiftStack as perfectly foreseeable future engineering ideas.
Das said Nvidia prospects would be capable to use business storage from any vendor. The SwiftStack 1space engineering will form the “storage orchestration layer” that sits involving the compute and the storage to properly position the info so the AI stack operates optimally, Das said
“We are not a storage vendor. We do not intend to be a storage vendor. We are not in the company of providing storage in any form,” Das said. “We get the job done pretty closely with our storage partners. This acquisition is intended to further the integration involving different storage systems and the get the job done we do for AI.”
Nvidia partners with storage distributors these kinds of as Pure Storage, NetApp, Dell EMC and IBM. The storage distributors combine Nvidia GPUs into their arrays or sell the GPUs along with their storage in reference architectures.
Nvidia captivated to open resource tech
Das said Nvidia observed SwiftStack attractive since its software is dependent on open resource engineering. SwiftStack’s eponymous item-and file-dependent storage and info administration software is rooted in open resource OpenStack Swift. Das said Nvidia ideas to continue to get the job done with the SwiftStack team to progress and optimize the engineering and make it readily available through open resource avenues.
“The SwiftStack team is section of Nvidia now,” he said. “They’re super gifted. So, the innovation will continue to happen, and all that innovation will be upstreamed into the open resource SwiftStack. It will be readily available to any one.”
SwiftStack laid off an undisclosed selection of profits and advertising personnel in late 2019, but saved the engineering and assistance team intact according to president Joe Arnold. He attributed the layoffs to a change in profits aim from basic backup and archiving to AI, equipment mastering (ML) and info analytics use scenarios.
The SwiftStack 7. software update that emerged late previous calendar year took goal at analytics HPC, AI and ML use scenarios, these kinds of as autonomous vehicle applications that feed info to GPU-dependent servers. SwiftStack said at the time that it experienced labored with prospects to design and style clusters that could scale to tackle various petabytes of info and assistance throughput in excess of 100 GB for each next.
Das said Nvidia has been utilizing SwiftStack’s item storage engineering as perfectly as 1space. He said Nvidia’s inside get the job done on info science and AI applications experienced quickly confirmed the organization that accelerating the computer system shifts the bottleneck somewhere else, to the storage. That performed a variable in Nvidia’s acquisition of SwiftStack, he noted.
“We regarded a lengthy time ago that the way to assist the prospects is not just to offer them a GPU or a library but to assist them create the full stack, all the way from the GPU up to the applications on their own. If you search at Nvidia now, we spend most of our strength on the software for different types of AI applications,” Das said.
He said Nvidia would entirely assistance SwiftStack’s shopper base. SwiftStack promises it has all-around a hundred twenty five prospects. It solutions lineup integrated SwiftStack’s item storage software, ProxyFS file system for built-in file and item API accessibility, and 1space. SwiftStack’s software is intended to operate on commodity components on premises, and its 1space engineering can operate in the public cloud.
SwiftStack put in far more than eight yrs increasing its software’s capabilities because the firm’s 2011 founding. Das said Nvidia has no reason to sell the SwiftStack’s proprietary software since it does not compete head to head with other item storage vendors.
“Our philosophy right here at Nvidia is we are not making an attempt to compete with infrastructure distributors by providing some type of a stack that competes with other peoples’ stacks,” Das said. “Our aim is merely to make people today prosperous with AI. We believe, if that transpires, all people wins, like Nvidia, since we believe GPUs are the most effective platform for AI.”