Vast Data Inc. has added block storage capabilities and an Apache Kafka-compatible event streaming service to its flagship Vast DataStore offering, transforming it into what it says is the industry’s first “fully unified” data platform for artificial intelligence workloads.

With today’s announcements, Vast says it has fulfilled its vision of creating a “universal storage platform” that can linearly scale parallel data access across every type of data format, including file, object, block, table and streaming data. By making all of these data types accessible through a single storage platform, it’s possible for companies to cater to every kind of workload, including AI, through a unified storage architecture, without making any compromises in terms of performance, economics and scalability, the company said.

The company says it’s satisfying enterprise’s demands for “multiprotocol storage platforms,” which are defined by Gartner Inc. as being able to support multiple storage access protocols to address the growing needs of businesses. Such platforms must be versatile, enabling data to be stored and accessed via multiple protocols, such as the Network File System, Server Message Block, block and object. By providing this versatility, Vast says, it can support streamlined integration with a multitude of information technology environments, while ensuring it meets the storage requirements of every kind of application.

Vast DataStore now supports a comprehensive range of virtualization platforms, including VMware, Hyper-V and other hypervisors. Features such as multitenancy and QoS make it simple for teams to isolate workloads and simplify storage management across hundreds of thousands of virtual machines, the company says.

Moreover, the addition of block storage means it can support persistent storage for Kubernetes- and OpenShift-based containerized applications, including transactional databases and stateful microservices. As a result, Vast says, it can consolidate siloed storage infrastructure into a single platform with support for enterprise-grade features such as snapshots, multitenancy, quality of service, replication, encryption and role-based access controls.

NAND Research Inc. Chief Analyst Steve McDowell told SiliconANGLE it’s encouraging to see Vast following a broader industry shift to unified storage, where platforms support every kind of data format. It gives it the ability to service all major storage workloads, without customers having to segregate things based on the underlying storage protocol.

However, he said its block storage capabilities cannot yet be considered complete, as it currently only supports NVME-over-Ethernet, lacking the ability to handle more traditional storage array network protocols such as Fiber Channel and iSCSI. Moreover, its NVME solution doesn’t yet support remote direct memory access adapters, which means its performance might be limited compared with some of its competitors.

“These capabilities, plus VMware vVols and Kubernetes CSI support, are all road-mapped features that should land sometime down the road, so as long as that road doesn’t stretch too far into the future, it shouldn’t keep customers from evaluating the solution,” McDowel said. “The pieces it’s currently lacking should be available soon.”

The more exciting addition announced today is the availability of Vast’s new Event Broker, a data streaming messenger service based on Apache Kafka. According to the company, this means its platform can now support real-time analytics, event-driven workflows and AI and machine learning pipelines, while offering expanded Structured Query Language capabilities, simplified management and improved observability.

McDowell said this is another example of how VAST continues to lead the industry in terms of integrating data management capabilities directly into its data infrastructure. “It’s a great approach that simplifies implementing a solution for IT teams, while also delivering a substantial performance boost,” he said.

What it means is that, instead of data having to traverse between external servers and take a performance hit as it’s copied between nodes, it can be streamed directly into Vast’s DataStore. It’s much more efficient than any architecture that requires data movement, McDowell explained.

“It will prove valuable to enterprises with performance data needs, such as advanced analytics and front-end to AI pipelines,” the analyst said. “I really like this approach, and the choice of building a Kafka-compatible solution is the right one as it has emerged as the de facto industry standard for data streaming, so it’s easy to deploy.”

Image: SiliconANGLE/Freepik AI Suite