Panzura LLC is updating its CloudFS hybrid cloud file platform with features aimed at lowering storage infrastructure costs, simplifying operations and preparing enterprise file data for emerging agentic artificial intelligence workloads.

Released today, Version 8.7 focuses on reducing the operational burden of large-scale file environments while giving information technology teams greater control over distributed file systems that support global collaboration, the company said. The target market is organizations whose intellectual property resides in large project files, initially architecture, engineering and construction firms.

“The focus is on helping customers control infrastructure costs, simplify management and probe the environment for agentic AI buffers,” said Chief Product Officer Karthik Ramamurthy. The update reduces storage requirements while increasing automation and self-service capabilities.

Preparing for AI workloads

The release responds to the fact that many organizations find data accessibility, not model capability, is the biggest barrier to deploying AI, Ramamurthy said. “Accessing data that matters is a real barrier,” he said.

The new version of CloudFS aims to address that problem by consolidating unstructured data, metadata and version history into a unified namespace that AI pipelines can access directly.

“CloudFS is basically a unified platform,” said Darrin Chapman, senior director of product management. “We govern, protect and prepare data for AI without requiring any kind of professional services support or ticketing.”

Adaptive Snapshot Retention is a new feature that reduces metadata overhead by allowing different snapshot policies to be applied across nodes in a distributed environment. Historically, organizations running distributed file systems had to retain identical snapshot histories across nodes, Panzura said. That requirement significantly increased metadata volumes and infrastructure costs as environments grew.

“With 8.7, we’re allowing you to keep the full history in the core, but a smaller history at the edge,” Chapman said. Reducing snapshot counts on edge nodes lowers metadata requirements, which reduces the need for additional storage space. “We’re starting to reduce that metadata footprint,” Chapman said.

More control for IT teams

CloudFS 8.7 also introduces a set of self-service administrative features intended to reduce operational friction and eliminate support bottlenecks.

File lock release allows administrators to clear locked files without opening support tickets, addressing a common collaboration problem that occurs when users disconnect while editing shared files.

Prewarm provisioning equips new nodes in a distributed environment to start with fully synchronized datasets so that “when teams arrive in a new location, the full localized dataset is already available,” Chapman said.

Those features are also designed to support future autonomous operations as AI systems increasingly interact with enterprise data environments. “Agentic systems simply can’t wait for a support ticket,” Chapman said.

New AI-powered user behavior intelligence is a security feature that analyzes user activity patterns to detect potential threats. As enterprises begin exposing large volumes of file data to AI systems, security monitoring is becoming increasingly critical, Chapman said.

The system analyzes behavioral patterns and flags abnormal activity, such as unusual download volumes or unexpected access requests. “We can detect and contain any kind of anomalies before they become a problem,” he said.

Though Panzura is initially targeting architecture, engineering and construction firms with the new release, the architecture is not limited to those sectors. Ramamurthy said manufacturing, life sciences and media organizations are future target markets.

As AI adoption accelerates, file systems will play an increasingly central role in enterprise AI architectures using graphics processing unit, he said. “A lot of the work we do on the file system side is about making sure data is accessible very close to the GPUs that are doing all the inferencing,” he said.

Image: SiliconANGLE/Dreamina