Scality’s Autonomous Data Infrastructure does agent-driven tiering and more

Object storage supplier Scality has launched an enterprise-focussed data infrastructure management product called ADI (Autonomous Data Infrastructure) to place data in four performance, cost, and protection storage tiers with policy-driven AI agent workers.

The background to this is that Scality believes there are too many, wide-ranging demands on an enterprise’s on-premises object storage system for human operators to manage on their own. AI workloads such as training, inference, multimodal agentic workflows, retrieval-augmented generation (RAG), Video Search and Summarization (VSS), and KV cache for distributed inference, have different requirements for throughput, latency, and data governance. A data storage system has to respond to power constraints, be cyber-resilient, meet sovereignty requirements and support massive scalability while not becoming unfeasably expensive and complex to operate

Scality CEO Jérôme Lecat said: “The AI era hasn’t just changed how enterprises use data, it has exposed how badly the old storage model was broken. Scality ADI isn’t just a faster object store. It’s a new operating model that autonomously aligns the right performance, protection, and economics to every workload, at every stage of the data lifecycle. That’s what it takes to keep GPUs productive, satisfy regulators and insurers, and maintain sovereign control, all at the same time, and at exabyte scale. We are not replacing what works. We are building what comes next.” 

Envisage a customer with Scality’s RING or RING XP object storage on-promises, supporting AWS, Azure and GCP clouds, and accompanied by Scality’s ARTESCA immutable backup object storage. This customer needs some data to be kept in a sovereign state, other data to be fed lightning fast to GPUs, with a hierarchy of caches to cope with HBM and CPU DRAM limits, and less-accessed data to be retained in tape storage, on-premises or in the cloud, all the while maintaining cyber-resiliency.

Scality says standard storage array control and management software can’t do this. Customers almost inevitably end up with separate storage products to meet these differing needs or try to force fit them into single-tier flash storage. What it’s proposing is a data management infrastructure layer that is powered by AI agents and oversees a multi-tier but single namespace (Scality) storage environment that provides the extremes of performance, cyber-resiliency, data sovereignty, power-consumption management and enables sysadmins to cope with massive scale.

It says ADI “aligns the right media, protection, and performance to each workload, so you can power AI, protect critical data, and scale sustainably, without choosing between them.”

ADI supports a policy-driven data lifecycle across 4 tiers:

1. Extreme performance tier with GPU-Direct, S3 over RDMA accessed TLC flash drives with <50μs latency,
2. Hot tier with QLC and future near-line SSDs accessed by S3 over RDMA with multi TB/s bandwidth,
3. Warm tier of future NL-SSD, and NL HDDs again accessed by S3 over RDMA,
4. Cold tier of tape and public cloud targets.

For effective power sustainability, it suggests 5 percent of data on tiers 1 and 2, 30 percent on the warm tier and the rest, 65 percent, on the cold tier. Scality says: “Real-time power telemetry gives infrastructure teams visibility into consumption at system, node, and workload levels, connecting performance decisions to actual data center constraints.” 

A Guardian component provides AI agent-assisted, operational intelligence. It “Observes system state and surfaces workload-aligned insights across predictive maintenance, platform health, power consumption, and cyberthreat detection.” The agents “handle expansion, healing, rebalancing, upgrades, and lifecycle workflows.”

Customers can integrate their own AI tools into ADI's operational workflows via MCP, “so the platform can be driven by a customer’s own AI stack, not just Scality’s.” 

We’re told “agents surface insights, humans (or your AI tooling) approve actions, and the platform executes within auditable policy bounds, reducing manual burden at scale.” ADI is not completely autonomous.

Scality's CORE5 cyber resilience offering ensures data remains immutable, recoverable, and auditable at every level. 

Scality ADI is available now through Scality’s global network of channel partners and strategic alliance partners. Its source code is available for inspection and governed contributions and it’s backed by outcome-based SLAs spanning availability, performance, protection posture, power consumption and operational efficiency. Get more information here.

Comment

ADI does not support Nvidia's STX KV Cache scheme, but we expect it will.