Ctera Networks Ltd. says it’s building a better data foundation for artificial intelligence systems, consisting of a new federated data architecture that eliminates the tradeoff between traditional file systems and object storage.

The startup’s new product is called Ctera Fusion Direct. Announced today, it unifies the two standalone storage domains into a single, high-performance data fabric that allows both structured and unstructured information to be accessed from the same location at unprecedented speeds.

According to Ctera Chief Executive Oded Nagel, enterprises historically have had to muddle their way through two totally disconnected storage worlds. On the one hand, they’re dealing with traditional network-attached storage systems that are optimized for human access. On the other hand, there’s object storage, which is primarily designed for machine-driven workloads.

Both storage mediums are common, but totally incompatible with one another. To share information across these systems, organizations have to build performance-limiting translation layers and accept that their data will have to be duplicated across multiple systems.

Ctera Fusion Direct creates a “unified fabric” that sits between NAS and object storage, enabling the two file types to coexist natively within the same, federated global namespace. In other words, they can sit in what is essentially the same storage environment, where data can be written as files and read as objects and vice versa. The new fabric eliminates the file-to-object conversion bottleneck, because there’s no more need to build translation gateways and proprietary chunking schemes, Nagel said.

Nagel explained that the main roadblock for enterprise AI is not the lack of data, but the struggle to use it effectively. He talked about a “historic divide” between data for human collaboration and data for machine analytics, which creates massive friction and complexity for AI systems that need access to both.

“We’ve collapsed those silos,” he said. “This fundamental shift in architecture unlocks the full potential of a customer’s data, providing the unified data platform necessary to not just compete, but to lead in an economy increasingly driven by machine learning.”

Organizations will be able to attach their existing object storage buckets directly to the new data fabric and enable both humans and machines to access information spanning multiple cloud environments and edge locations, without any duplication, migration or modification. In such cases, the data within them will be presented to humans as traditional files, while machines will continue to access it via server message blocks as before, Nagel said.

Ctera Fusion Direct supports high-performance standards such as S3 over RDMA, enabling massive datasets to be streamed directly from object storage to file-based applications, removing the need for time-consuming local downloads first. In addition, it means AI clusters and GPUs will be able to read and write data at maximum bandwidth. All data remains in standard Amazon Web Services S3 buckets, ensuring organizations remain in full control of their sensitive information assets and intellectual property.

Frost & Sullivan analyst Karyn Price said Ctera’s data fabric is compelling, because numerous companies still struggle to tap the value hidden within their unstructured datasets. “File systems make data useful for people, but they often make it inaccessible to modern AI systems and analytics engines,” she said. “A platform that can fluently speak the languages of both, presenting data as files to humans and as objects to machines, removes a fundamental barrier to progress.”

Image: Ctera