Tag: Architecture

Infinidat has introduced a retrieval augmented generation (RAG) workflow architecture, offered as a consulting service to its storage clients, enabling them to incorporate current, private data from various company sources into artificial intelligence (AI) from any NFS storage within their organization. This development aligns with a trend where numerous storage companies are focusing on AI workloads and addressing RAG issues, particularly in generative AI (GenAI), which arise when the training data is incomplete, outdated, or lacks specialized information that can only be obtained from private data sources, such as within an organization or through expert knowledge. Organizations looking to develop GenAI typically subject a dataset to a training process where the AI learns to identify specific attributes for information or application triggers. These training processes often rely on generalized datasets that may become outdated or lack specialized or private information initially. This is a common scenario in AI projects within organizations that require ongoing updates, explained Infinidat's chief marketing officer, Eric Herzog. "Many organizations are leveraging generative AI projects with private data and are concerned about maintaining accuracy and avoiding errors," Herzog stated. "For example, a large enterprise producing vast amounts of data in various areas like sales, support, and operations aims to enhance performance linked to storage efficiency." Customers seeking accurate, real-time data can utilize AI to analyze details like component screws, types, suppliers, and other specifics, updating this information continuously. Infinidat now offers professional consulting services to enable customers to access data for RAG purposes from their own and other suppliers' NFS file storage systems. Herzog highlighted the assistance provided in configuring the storage system for rapid access to data and metadata for RAG purposes, emphasizing Infinidat's expertise in metadata and its architecture's "neural cache" in the InfuzeOS environment. Infinidat's arrays, available in all-flash or hybrid configurations, cater mainly to high-end enterprise and service provider clients. Their hardware products feature triple-active controllers and leverage a "neural cache" to optimize data placement, boasting a cache hit rate of over 90%. Infinidat's focus on RAG capabilities aligns with other storage vendors making strides in supporting customers embarking on AI initiatives. Pure Storage's CEO, Charlie Giancarlo, highlighted the company's AI initiatives at the Accelerate event, emphasizing storage write speed and availability. Similarly, NetApp unveiled a data management push for AI projects, introducing data classification for AI through its OnTap operating system at the annual Insight event.

Tag: Architecture

Infinidat gets in on the RAG act with workflow architecture offer

HONOR Presented Four-Layer AI Architecture And Google Partnership at VivaTech 2024

Quick Links

Resources

Support