Manuvir Das, Vice President of Enterprise Computing at Nvidia, explained to CRN, When text is converted to embeddings and stored, the size of these embeddings is significantly larger—potentially ten times bigger than the original text. This represents a massive data and storage opportunity that the market has yet to fully grasp.
At Dell Technologies World 2024, Jeff Clarke, Dell's Vice Chairman and COO, emphasized the crucial role of storage in AI systems, comparing it to the lungs that pump data. "Storage is essential to AI's functioning," Clarke stated during the event.
CR Howdyshell, CEO of Dell Titanium partner Advizex, highlighted the importance of high-quality storage in handling the vast amounts of data consumed by the latest compute stacks. He advised prioritizing storage considerations early in the planning process to meet the needs of technically savvy customers.
Dell's Partner First For Storage program, introduced in August of last year, incentivizes core sellers to close storage deals through channel partners. This approach has attracted new customers for Titanium partners like Advizex and Platinum partners like VirtuIT, with Dell targeting a $13 billion market in mid-range storage.
John Lee, CTO of VirtuIT, noted the potential for refreshing old infrastructure to prepare small and medium enterprises for generative AI. "Updating infrastructure now ensures readiness for AI implementations in the future," he said.
Arthur Lewis, Dell's President of Infrastructure Solutions Group, underscored the growing storage demands of AI workloads, which require 300 times more data throughput than traditional compute tasks. He predicted that AI machines would demand 27 quettaflops of computing power by the decade's end.
Manuvir Das from Nvidia elaborated on the future of AI, which involves processing more complex, multi-modal data forms. "The ecosystem is evolving to handle and interpret rich data, such as videos, with advanced models," Das said.
Das also highlighted two key insights about AI data storage: embeddings, necessary for efficient data retrieval, are much larger than the original data, creating significant storage needs.
Dell's software-defined PowerScale, designed with AI in mind, offers high performance with GPU-direct technology. Additionally, Dell introduced the PowerScale F910, a high-performance file solution for unstructured data, featuring hardware upgrades like DDR5 and PCIe Gen5.
Looking ahead, Dell's Project Lightning, expected next year, promises to revolutionize unstructured storage for AI with performance and throughput increases of 20x and 18.5x, respectively. Lewis explained that as AI models become more specialized, the need for efficient storage access will increase.
Howdyshell expressed enthusiasm for the new PowerStore with 5-to-1 compression, highlighting the significant opportunity for combining rapid compute capabilities with robust storage solutions. "Customers need fast compute and reliable storage, making this a substantial opportunity," he concluded.
O’Ryan Johnson, a veteran news reporter at CRN, continues to cover the data center beat and seeks feedback from channel partners to enhance his coverage. He can be reached at ojohnson@thechannelcompany.com.
Post a Comment