The success of AI, especially generative AI the past year or so, hinges on one often-overlooked element: In this case, it's storage performance. Imagine a self-driving car struggling to access real-time traffic data due to sluggish storage. Its response time would lag, potentially leading to dangerous situations. Similarly, a generative AI attempting to craft a realistic portrait might produce blurry, pixelated results if its access to high-resolution images is hampered.
In these scenarios, the limitations of storage become the limitations of the AI itself, holding back its potential for transformative applications. By prioritizing storage performance, we unlock the full potential of AI and generative AI. Faster access to data means faster learning, quicker training times and more accurate and sophisticated models.
To accelerate traditional AI and generative AI, Dell announced its latest storage advancements. The purpose of these innovations is to address the critical bottleneck of data access and pave the way for faster model training, inference and overall improved AI outcomes.
These advancements include Dell PowerScale, Dell PowerScale OneFS software, a partnership with NVIDIA and Dell APEX File Storage for Microsoft Azure.
At the heart of Dell's AI storage lies PowerScale, the company's flagship scale-out file system. The latest PowerScale OneFS software enhancements streamline the AI workflow from data preparation to model training and inference.
OneFS now boasts improved parallel processing capabilities for faster data ingestion and preprocessing for AI models. This translates to quicker model training times and reduced overall development cycles. There are also OneFS optimizations for small file access that accelerate model training and fine-tuning. This is particularly beneficial for GenAI applications, which often involve working with large datasets of small files.
Additionally, OneFS's new caching and data placement algorithms minimize latency during model inference for faster and more responsive AI applications. This is important for real-time applications like fraud detection or anomaly recognition.
"Dell continues to design its file and object storage solutions with the customer in mind, allowing them to modernize their infrastructure with an AI-ready foundation to support the most intensive workloads, such as generative AI," said Scott Sinclair, Practice Director, Enterprise Strategy Group. "With PowerScale's latest performance advances, customers can improve AI model training and fine-tuning to make better business decisions based on their data."
And, to further amplify its storage advancements, Dell announced a strategic partnership with NVIDIA, provider of AI computing platforms. Dell's PowerScale systems are now validated for use with the NVIDIA DGX SuperPOD, a pre-configured, exascale-class AI infrastructure designed for demanding workloads. This collaboration ensures seamless integration between Dell's storage solutions and NVIDIA's AI computing platforms. This allows organizations to deploy and scale their AI and GenAI initiatives with confidence, knowing they have the best-in-class tools at their disposal.
With that said, Dell's storage innovations extend beyond raw performance, addressing the critical aspects of scalability, security and sustainability. PowerScale's flexible architecture scales from small deployments to petabyte-scale environments, accommodating the growing data demands of AI and GenAI applications.
On top of that, Dell is helping businesses process data anywhere by expanding its Dell APEX Storage for Public Cloud portfolio with a new file storage solution for Microsoft Azure. Dell APEX File Storage delivers enterprise-grade performance and management for demanding AI and ML applications, like Azure OpenAI Service and Azure AI Vision, directly within Microsoft's cloud environment.
This new offering goes beyond Azure, enabling organizations to leverage cloud-native AI and GenAI workflows on data across any public cloud or on-premises location with improved data access and movement. Through Dell's growing Databricks partnership, customers can choose from a range of large language models and retrain them with their own data stored in Dell APEX File Storage, regardless of their chosen cloud platform.
This collaboration furthers Dell's commitment to data management ecosystems that accelerate AI and analytics initiatives, further bolstered by its planned open, modern data lakehouse that integrates Starburst's analytics software with Dell's powerful hardware and storage solutions.
“Customers are relying on us to continually push the boundaries of storage innovation, including removing data access bottlenecks that limit the throughput and scalability of compute-intensive applications,” said Arthur Lewis, President, Infrastructure Solutions Group, Dell Technologies. “We are addressing these needs by delivering fast, efficient and secure access to data and turning it into a proverbial goldmine of AI and GenAI possibilities.”
Edited by
Alex Passett
