오늘날 AI 모델은 새로운 독창적인 데이터를 생산하는 동시에 텍스트, 이미지, 오디오, 비디오 등을 처리하는 연속적인 데이터 생성과 소비의 반복을 기반으로 작동한다. AI 기술이 발전함에 따라 데이터 스토리지 시스템은 대량의 데이터를 관리하면서 방대하고 복잡한 모델에 요구되는 컴퓨팅 로드와 속도를 지원하는 것이 중요하다.

▲Western Digital's 'AI Data Cycle' product line / (Photo: WD)
Announcing storage solutions for AI workloads
Today’s AI models operate on the repetition of continuous data generation and consumption, processing text, images, audio, video, etc. while producing new and unique data. As AI technology advances, it is important for data storage systems to manage large amounts of data while supporting the computing load and speed required for large and complex models.
Storage solutions company Western Digital (WD) on the 11th unveiled a six-step AI data cycle framework that defines the ideal storage configuration for large-scale AI workloads. Along with this, the 'Ultrastar DC SN861 SSD', 'Ultrastar DC SN655 SSD', and 'Ultrastar DC HC690 UltraSMR HDD' were also announced.
The new AI Data Cycle Framework helps customers plan and build advanced storage infrastructure to maximize their investments in artificial intelligence (AI), improve efficiency, and reduce the total cost of ownership (TCO) of AI workflows.
Western Digital has established a roadmap of flash memory and HDD products and technologies to meet the storage requirements needed for key stages of the AI data cycle, with new products including high-performance PCIe Gen5 SSDs to support AI training and inference, high-capacity 64TB SSDs for high-speed AI data lakes, and ePMR UltraSMR 32TB HDDs that provide cost-effective large-scale storage.
The Ultrastar DC SN861 SSD is Western Digital’s first enterprise-grade PCIe Gen 5.0 solution, delivering exceptional random read performance and best-in-class power efficiency for AI workloads. With up to 16TB of capacity, it delivers up to 3x faster random read performance than the previous generation, with ultra-low latency and responsiveness for LLM training and inference, and AI service applications.
It also enables lower IOPS/W through a lower power profile, supporting a reduced overall TCO. Enhanced PCIe Gen5 bandwidth addresses the AI market’s growing need for high-speed accelerated computing and low latency for compute-intensive AI environments.
Designed for mission-critical workloads, the Ultrastar DC SN861 SSD features NVMe® 2.0 and OCP 2.0 support, 1 and 3 DWPD (Drive Writes per Day), and a five-year limited warranty. The Ultrastar DC SN861 SSD E1.S model is sampling now, with the U.2 model expected to sample in June and mass ship in the third quarter of this year. Additional information on the E1.S and E3.S form factor models will be announced later this year.
The newly expanded lineup of enterprise-grade Ultrastar DC SN655 SSDs is designed for storage-intensive applications. New U.3 SSD options support capacities up to 64TB, delivering enhanced performance and capacity for AI data prep and faster, larger data lakes. The new U.3 SSD model in the Ultrastar DC SN655 SSD lineup is currently sampling, with additional information to be announced alongside mass shipments later this year.
“Generative AI is the next big thing, and storage will play a critical role in it,” said Ed Burns, research director at IDC. “Generative AI is expected to significantly change what it means to mean storage, as storage and access to data become increasingly important to the speed, efficiency, and accuracy of AI models, especially as larger, higher-quality datasets become more widespread.”
“As AI technologies take root across nearly every industry, storage has become an increasingly critical component of the AI technology stack,” said Rob Soderbery, executive vice president and general manager of Western Digital’s flash business unit. “Our new AI Data Cycle Framework is designed to help customers build storage infrastructure that delivers tangible impact on the performance, scalability and utilization of AI applications.”