AiRE platform
The data set is the unit of configuration.
AiRE is a proprietary software stack on a storage-optimised UNIX operating system, built around a 128-bit pooled file system. Each data set carries its own profile, encryption, reduction and protection — inside the same unit.
- Unified block and file
- Copy-on-write immutable snapshots
- In-line dedup and compression
- Self-healing integrity
At a glance
- File system
- 128-bit, pooled, copy-on-write
- Protocols
- Block and file served from the same unit
- Deduplication
- In-line, per data set, SHA256 checksums
- Compression
- In-line, LZ4 or GZIP9 per data set
- Protection
- RAID-Z or mirror topologies
- Integrity
- End-to-end checksums to the root node
- Snapshots
- Point-in-time, immutable, instant clones
- Replication
- Block-level incremental, deltas only
Figures and mechanisms as published by STORViX.
What AiRE is
Software first, hardware second.
AiRE is STORViX's proprietary software stack, built on a storage-optimised UNIX operating system with a robust file system and software-defined data services. It runs on the AiRE Unified Data Platform appliance or as a virtual instance; either deployment is called an AiRE instance.
The design intent is drawn from how biological systems adapt: reproducing a DNA sequence encoded in binary, which analyses and monitors the storage unit's functionality. The platform is meant to learn from its own telemetry and adjust to the workload rather than being tuned once at install and left alone.
In practice, the consequence that matters commercially is this — because configuration lives at the data set rather than at the array, a single unit can carry a latency-sensitive virtualisation workload and a compressed long-term archive at the same time, each with its own encryption, reduction and protection settings.
Architecture, top to bottom
Cloud services
CloudSight telemetry · AutoPILOT · CoPILOT Connect · fleet management
Data services
Encryption · deduplication · compression · snapshots · replication
Pooled file system
128-bit, copy-on-write · unlimited data sets across pools · variable block sizes
Storage-optimised UNIX OS
I/O path management · cache behaviour · RAID-Z and mirror topologies
Hardware or virtual instance
SPU with single or dual controllers · disk packs across tiers 0–3 · DEU expansion
Each layer is described in STORViX's published technical material. The stacking is an editorial simplification for orientation, not a statement about internal module boundaries.
What that means in practice
Four workloads that would normally need three systems.
Because configuration lives at the data set rather than the array, unlike workloads share one unit without sharing one compromise setting.
One AiRE instancephysical appliance or virtual
Pool A
Accelerated all-flash · NVMe
- vm-datastorePerformance
- Small block size
- LZ4 compression
- Encrypted
- Mirror
- db-volumesPerformance
- Low latency path
- No deduplication
- Encrypted
- Mirror
Pool B
Hybrid-flash · HDD + SSD
- imaging-archiveOptimization
- Large block size
- GZIP9 compression
- Deduplicated
- RAID-Z
- compliance-recordsResilience
- Immutable snapshots
- Vault replication
- Encrypted
- RAID-Z
An unlimited number of data sets across pools, all inside the same unit
Core capabilities
What the platform actually does.
Named mechanisms and algorithms, so an evaluator can check them rather than take them on trust.
Pooled storage, not fixed parameters
Conventional arrays fix their parameters when the system is configured. Changing them later means wiping the array or buying another one. AiRE creates an unlimited number of data sets across pools inside the same unit, each with its own settings.
- Hundreds of parameters settable per data set
- Profiles and policies instead of manual tuning
- Variable block sizes for volumes and file systems
- No forklift reconfiguration
Three workload profiles
Performance prioritises low latency and throughput. Resilience emphasises redundancy, replication and protection. Optimisation minimises system load and maximises usable space. Policies then adjust I/O path management, cache behaviour and block size to match.
- Performance — virtualisation and latency-sensitive work
- Resilience — mission-critical data
- Optimisation — capacity efficiency and cost
- Policies customisable from best-practice defaults
Data reduction applied in line
Deduplication and compression happen during the write, not as a post-process pass. Deduplication is granular to the data set, so reduction effort is spent where it pays.
- In-line deduplication with 256-bit checksums (SHA256)
- In-line compression: LZ4 for speed, GZIP9 for density
- Per-data-set control rather than array-wide
- Customer-reported compression factors of 2× and 3×
Integrity that repairs itself
Every block is checksummed, and the checksum is stored in a reference rather than in the block. That chain extends to the root node. On read, a mismatch triggers self-repair from redundant copies.
- End-to-end checksums through the whole hierarchy
- Automatic repair of corrupted blocks
- RAID-Z parity or mirror topologies
- Protection against silent data corruption and decay
Recovery through immutable snapshots
A copy-on-write transactional model produces point-in-time snapshots that cannot be altered once taken — which is what makes them useful against ransomware. Clones are instant and consume no additional space.
- Immutable point-in-time snapshots
- Instant clones with no extra capacity
- Block-level incremental replication of deltas only
- Minimal bandwidth for off-site copies
Off-site into Vault
Replicate or mirror to STORViX-operated data centres in Sweden, Italy and other EU countries. Recovery is over the network, or by having physical drives shipped.
- SHA AES 256 encryption at rest
- FIDO2 multi-factor authentication on physical drives
- Zero-knowledge privacy policy and DPA
- Geographically separated EU facilities
Security and resilience
Controls that are demonstrable, not asserted.
STORViX positions AiRE as engineered for strict NIS2 requirements. The controls below are technical facts; whether they satisfy a given obligation is a legal assessment for your organisation and its advisers.
Immutable snapshots
Copy-on-write, unalterable once taken — the basis of ransomware recovery
Per-data-set encryption
Applied to each data set individually rather than array-wide
End-to-end integrity
Checksums to the root node, with automatic self-repair on mismatch
Two-factor authentication
Administrative access protected by 2FA
Anomaly detection
Deviations surfaced from CloudSight telemetry as they occur
EU data residency
Vault replicas held in Sweden, Italy and other EU countries
Zero-knowledge Vault
Encrypted at rest, FIDO2 MFA on physical drives
A technical control is not a compliance certification. STORViX describes AiRE as engineered to exceed NIS2 standards; this page does not claim certification on your behalf.
Operational benefits
The platform administers a good deal of itself.
Telemetry goes up, decisions come back down. The point is fewer administrator hours per usable terabyte, not novelty.
CloudSight
AiRE sends telemetry — performance metrics, configuration, health, workload types, capacity utilisation — to the cloud, where machine learning identifies sub-optimisation, errors and emerging problems, and raises alerts to both the customer and the SmartCARE team.
- Real-time analysis
- Prescriptive reporting
- Predictive capacity forecasting
AutoPILOT
An AI-driven system that performs optimisations, fixes and administration autonomously, continuously analysing telemetry. When a problem exceeds what it can resolve, it flags the issue for human intervention rather than failing silently.
- Automated profile recognition
- Continuous policy enforcement
- Escalation to a human on unresolved issues
CoPILOT Connect
An always-available remote assistance channel over the two-way cloud connection, so the SmartCARE team can apply fixes and optimisations without waiting for an access window.
- Two-way cloud connection
- Remote corrective action
- Available across all support tiers
Centralised fleet management
Software updates, configuration changes and commands are distributed to as many AiRE instances as needed at once, keeping a whole estate consistent rather than drifting instance by instance.
- Fleet-wide updates
- Consistent configuration
- One console for many instances
AI and data-intensive workloads
Tiering that suits a pipeline, not just a benchmark.
AI work is rarely uniform. Training sets need throughput, checkpoints need durable low-latency writes, and the bulk of the corpus is cold between runs. Buying all-flash for all of it is how storage budgets get consumed by data nobody is reading.
Because AiRE applies profiles per data set, the active working set can sit on an accelerated all-flash pool while the corpus sits compressed and deduplicated on a hybrid pool — in the same unit, under one interface, without a separate archive product.
STORViX has been part of the NVIDIA Inception programme since 2022 and works with Oracle and AWS. Specific throughput figures for a given model and pipeline should come from a proof of concept against your data, not from a datasheet.
Deployment flexibility and hardware
System processing unit
A 4U chassis holding the SPU, configurable with a single controller or with dual controllers (A and B). All-flash configurations are always dual-controller.
- 4U base unit
- Single or dual controller
- GEN 4 is the current generation
Disk packs across tiers 0–3
Four disk pack types cover the cost/performance range: Hybrid Flash, Accelerated Hybrid Flash, All-Flash and Accelerated All-Flash — the last using a flash tier for I/O path metadata.
- HF — HDD and SSD, balanced cost
- AHF — higher-performance SSD and HDD blend
- AF — all-SSD including NVMe
- AAF — flash tier for I/O path metadata
Capacity that grows in 4U steps
A base unit alone provides 32–501 TiB net. Adding one disk expansion unit reaches 71 TiB–1,002 TiB; two reach 370 TiB–1.5 PiB. A full datacentre cabinet reaches up to 20 PiB.
- Base unit (4U): 32–501 TiB net
- Base + 1 DEU (8U): 71 TiB–1,002 TiB net
- Base + 2 DEU (12U): 370 TiB–1.5 PiB net
- Up to 20 PiB per cabinet
A seven-to-ten-year lifecycle
Disk pack configurations can change as demand changes, so performance and capacity scale without replacing the chassis. That is what keeps the hardware in service well past a conventional three-year refresh.
- No forced three-year refresh
- Modular expansion
- Lower TCO and embodied carbon
The hardware
What you actually rack.
A 4U system processing unit, and disk expansion units in 4U steps. Drawn to scale, so the difference between configurations is the difference you would see in the cabinet.
The unit itself
Technical illustration — not a photograph
Drawn to the true proportions of a 4U 19-inch chassis, from STORViX's published specifications: 4U form factor, dual controller, front-loading LFF SAS drive carriers. The drive-bay layout is indicative — a conventional 24-bay arrangement — and the bezel styling and badging are deliberately generic. STORViX has not supplied product photography or industrial-design artwork.
Fully expanded, in a cabinet
Technical illustration — not a photograph
A fully expanded system — one 4U processing unit and the maximum of two disk expansion units — drawn to scale in a standard 42U cabinet. The 4U form factor and the two-DEU maximum are from STORViX's published specifications; the 42U cabinet is the conventional reference rather than a STORViX product. Bezel styling and badging are generic, since no product photography or industrial-design artwork has been supplied.
32–501 TiB
Net capacity from a single 4U base unit
STORViX published specification
1.5 PiB
Net capacity at 12U, with two disk expansion units
STORViX published specification
20 PiB
Per datacentre cabinet
STORViX published specification
256-bit
Checksum strength used for in-line deduplication
STORViX published specification
Questions
Frequently asked
- Can I buy the hardware without the AiRE software?
- No. STORViX's position is that the two are designed together: AiRE is available as a virtual instance on its own, but the AiRE Unified Data Platform hardware is not sold without it.
- How large is one AiRE Unified Data Platform unit?
- The base unit without a disk expansion unit is 4U, offering 32 TiB to 501 TiB net depending on disk pack configuration. Each additional DEU adds 4U: one DEU gives 71 TiB–1,002 TiB, two give 370 TiB–1.5 PiB. Up to 20 PiB is achievable in a datacentre cabinet.
- What disk types are available?
- Four. Hybrid Flash blends HDDs and SSDs for balanced cost. Accelerated Hybrid Flash adds higher-performance SSDs. All-Flash is entirely SSD including NVMe. Accelerated All-Flash adds a flash tier dedicated to I/O path metadata for the most demanding workloads.
- Does AiRE serve block and file at the same time?
- Yes. AiRE manages both block and file protocols within the same unit, presented through a single interface, so a virtualisation datastore and a file share can live on one system with different data set settings.
- Which hypervisors and virtualisation platforms does AiRE work with?
- AiRE is hypervisor-neutral. It presents NFS and SMB for file and Fibre Channel and iSCSI for block, so any virtualisation platform that consumes those protocols — VMware, Proxmox, Hyper-V, KVM, OpenStack, XCP-ng — can use it as a datastore. STORViX's published customer material describes an AiRE system running as primary storage for a VMware cluster, and STORViX partners with Vates for XCP-ng where a fully European stack is the requirement. Formal certification or validation under a specific hypervisor vendor's programme is a separate question — ask STORViX for the current status rather than assuming it from this page.
- How does AiRE protect against ransomware specifically?
- Snapshots use a copy-on-write transactional model and are immutable once taken, so an attacker who encrypts live data cannot alter the snapshot history. Recovery is a revert to a known-good point. Encryption is applied per data set, and two-factor authentication protects administrative access.
- What happens to my data if a drive fails?
- AiRE detects the faulty disk and activates a spare. Data is reconstructed from RAID-Z parity or a mirror. Separately, end-to-end checksums mean a block that has silently decayed is detected on read and repaired from a redundant copy, rather than being returned as if it were intact.
- Is telemetry mandatory?
- CloudSight telemetry is what powers predictive capacity forecasting, AI-driven alerting and proactive SmartCARE support. Deployment questions for air-gapped or restricted environments should go to an architect, as the constraints differ per environment.
Put AiRE in front of your own workload.
An architecture conversation is more useful than a datasheet. Bring your capacity plan, your protection requirements and the workloads you are consolidating.