待翻譯：Supermicro alliance tackles the storage bottlenecks holding back enterprise AI

AI 服務暫時不可用，以下為來源正文，待恢復後補全翻譯。

The conversation around storage modernization used to revolve around capacity. Today, the dialogue centers on the role of storage as a strategic requirement for AI readiness. This doesn’t mean that storage’s transformation is simple. Many organizations are seeking to build AI on top of storage architectures that were designed before the AI era. These legacy environments are marked by fragmented silos and aging systems that complicate delivery of AI initiatives. To resolve this dilemma, organizations are adopting storage infrastructure that can provide critical elements of performance, flexibility and efficiency, according to Allen Liu (pictured, right), senior product manager of architecture solutions at Super Micro Computer Inc. “For Supermicro, what we want to do is … keep the performance and scale,” Liu said. “Second, we want to keep flexibility and composability so that makes sure the module can be used to assemble your solution. And the third is the efficiency. The efficiency is not only about performance, it’s not only about capacity, it’s also about usability, the TCO and all our solutions that make things all together work.” Liu spoke with theCUBE Research’s Rob Strechay for the Supermicro Open Storage Summit interview series, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. He was joined by Molly Presley (left), chief marketing officer of Hammerspace Inc., and Praveen Midha (center), director of enterprise SSD product management at Sandisk Corp. They discussed how organizations are modernizing storage environments, consolidating infrastructure and building foundations capable of supporting both current and future AI workloads. (* Disclosure below.) Software layers for storage modernization The process of making storage systems work together to enable AI readiness has been a central element in Supermicro’s partnership with Hammerspace and Sandisk. For Hammerspace, this has meant delivering a software solution that can channel data rapidly to AI models. “What we’re really looking for at Hammerspace is, ‘How do we take our data and our data infrastructure and make it available to your AI initiatives?’” Presley told theCUBE. “We are a software layer that you can lay on your existing infrastructure, use it in the cloud and the data center, at the edge, and we activate the data where it sits. You don’t have to do a big copy/migrate into a new AI storage system. You have a unified AI-ready dataset because we’ve ingested it into our metadata and we’re presenting it to your models, to your workflows quickly.” Channeling data to AI models and workflows is just part of the challenge for enterprises. There is an immediate need to keep pricey GPU systems processing data at maximum capacity, and Sandisk has developed quad-level cell, or QLC, solid-state drives that can improve capacity and efficiency across AI storage environments. “How do you feed the beast, how do you make sure that you keep these GPUs optimized and utilized if the average is 30% to 50% of utilization?” Midha said. “Our technology has to be something that can scale from a capacity perspective, which is where QLC comes in. We have a very strong offering in stage two for lower latency, higher performing media than just hard disk drives in stage one. As we go towards stage three, we are inside the GPU ring, we are really attached right next to the GPU.” Building new standards Storage’s evolution as a key part of the AI picture has been accompanied by the adoption of industry standards to facilitate the smoother movement of data. Supermicro has adopted protocols, such as NVMe over Fabrics, which extends local ultra-low latency NVMe storage across data center networks via networking technologies such as Remote Direct Memory Access. Standards such as these represent a key development in storage deployment, according to Presley. “The standards that are being developed right now are really important to talk about,” she said. “Some of them have been around for a bit, like NVMe over Fabrics and networks. A lot of standardization is critical if you want to work across these flexible environments, different vendors, different locations.” During theCUBE’s panel discussion, Midha presented slides that documented how the use of AI reasoning is dramatically expanding key-value (KV) cache sizes. Flash storage technology can facilitate more efficient retrieval of archived KV cache data from various tiers to speed up response times for new requests, according to Midha. “There’s a lot of excitement and a lot of opportunity for flash to make a meaningful contribution to this architecture,” he noted. “Flash is actually now a part of the memory tier.” The growing adoption of AI has led the storage industry to play a greater role in key tasks such as data serving, orchestration and efficiency. Supermicro’s partnership with Hammerspace and Sandisk highlights how the three companies are combining their strengths to enable new AI architectures. “Our partner, Hammerspace, they are a very good player in this [data orchestration] layer,” Liu told theCUBE. “The storage media layer, like Sandisk, is the place to hold data and to serve the data. The data needs to be served efficiently. Put all together, Supermicro is the infrastructure foundation. We put everything – data plus our own building blocks – to form the foundation for the data service.” Stay tuned for the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Supermicro Open Storage Summit interview series. (* Disclosure: TheCUBE is a paid media partner for the Supermicro Open Storage Summit interview series. Neither Supermicro, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.) Photo: SiliconANGLE A message from John Furrier, co-founder of SiliconANGLE: Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities. 15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more 11.4k+ theCUBE alumni — Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network Are you an AWS customer? Support SiliconANGLE financially by buying your AWS services from our Marketplace portal page and links: https://siliconangle.com/aws-marketplace/ About SiliconANGLE Media