HPE has made the 10.6.0 software release for the Alletra Storage MP B10000 generally available, landing inside the Q3 2026 window the company set when it previewed the release in May. HPE is also calling it Release 6 in its channel materials. The update takes the B10000’s disaggregated block-and-file architecture from four controller nodes to six, adds an agent-based support automation layer, folds real-time ransomware detection for both protocols into the array itself, and raises the platform’s capacity guarantee. HPE’s pitch is that unified storage has usually meant separate block and file products sharing a management pane, and that the B10000 consolidates the two on one shared-everything design where compute and capacity scale independently.

Six-Node Scale-Out and Workload Scope

The B10000 decouples controller compute from the capacity shelves behind it, so an organization can add performance or capacity separately as I/O profiles and footprints diverge, rather than buying both at once, as a dual-controller array requires. With 10.6.0, a cluster can grow from a single node to six in single-node increments. HPE puts the gain at a 50 percent performance increase over the previous four-node maximum, and says a six-node cluster can survive two simultaneous node failures without a service interruption. Customers that started on a switchless two-node configuration can also convert to a switched cluster without disruption, so a small deployment can grow into a larger cluster without a rebuild. Against a traditional scale-up array, HPE cites 40 percent TCO savings and a 45 percent reduction in energy consumption for the scale-out design, figures that come from HPE’s own substantiation rather than third-party testing.

The release also updates the HPE StoreMore Guarantee that backs the platform’s data reduction. HPE says the guaranteed effective capacity ratio moves to 5:1, up from 4:1, so a buyer sizing a B10000 can plan on more usable capacity per raw terabyte than before. As with any vendor guarantee, the ratio is workload-dependent, and the terms sit in HPE’s program documentation rather than the release itself.

HPE is positioning the platform for standard enterprise block workloads alongside adjacent unstructured file services, and 10.6.0 expands the file side with greater file capacity and snapshot support. HPE names SAP HANA and containerized applications as the workloads that need both protocols on the same system. It is not intended to replace the Alletra Storage MP X10000, which remains the company’s object platform for high-bandwidth AI, machine learning, EDA, and large-scale media pipelines. The aim of 10.6.0 is to consolidate general enterprise block and file data into a single system so that the two no longer reside in separate operational silos.

Agentic Support Automation

The more novel piece is what HPE calls agent-based support automation. Instead of threshold alerts that fire only after a metric crosses a fixed limit, the release deploys a set of specialized agents for detection, analysis, recommendation, and remediation that run continuously against the array’s telemetry. HPE says the system processes billions of log entries and sensor metrics a day, using GPU-accelerated inference to spot behavioral drift and correlate it with the workload responsible. The example HPE gives is uneven capacity exhaustion, where aggregate utilization looks healthy while one process quietly consumes the remaining headroom; the agents are meant to catch the pattern early and provide the administrator with a recommended fix, or, in some cases, carry out the remediation before an outage. HPE’s channel notice puts a qualifier on that last step: the remediation agent implements fixes with approval, so the autonomous path still runs through an administrator. Fleet telemetry and control stay in HPE’s Data Services Cloud Console, the GreenLake management plane for the Alletra line.

Cyber Resilience Built Into the Array

Data protection in 10.6.0 is part of the platform, not a downstream target. The B10000 now runs real-time ransomware detection natively on both the block and file access paths, and HPE says its Cybersecurity Center of Excellence validated the detection against more than 100 of the most common ransomware strains. The release adds continuous security posture monitoring, which checks the array’s security configuration in the background and flags drift for audits, and it protects immutable snapshot schedules from being altered. Immutable snapshots, integration with SIEM and XDR tooling, and role-based access controls with MFA, FIPS 140-3 compliance, and DISA-approved STIG hardening guidance round out the on-array controls, mapped to the NIST Cybersecurity Framework we walked through in our B10000 cyber resilience deep dive.

That array-level detection feeds a layered design. HPE Zerto Software supplies VM-level inline detection and continuous data protection for application recovery, and the release adds backup APIs plus integrated application-consistent backup to HPE StoreOnce for immutable retention, without an ISV in the path. Spreading detection, replication, and isolated retention across separate tiers is how HPE argues the platform avoids a single point of recovery failure during an attack.

The timing aligns with a run of third-party recognition that HPE is leaning on for the B10000. HPE cites IDC data calling it the fastest-growing all-flash block storage array, and HPE landed in the Leaders quadrant of the 2026 Gartner Magic Quadrant for Enterprise Storage. What 10.6.0 adds to that story is the ability to scale a unified block-and-file system out to six nodes, monitor it with an agent framework rather than a page of thresholds, and keep ransomware detection on the array instead of bolting it on afterward. For enterprise IT shops trying to collapse a general-purpose block array and a file server onto one platform, that is a more concrete consolidation path than the shared-console version of unified storage.

HPE Alletra Storage MP B10000

The post HPE Alletra Storage MP B10000 10.6.0 Arrives With Six-Node Scale-Out and Agentic Support Automation appeared first on StorageReview.com.

The Linear Tape-Open (LTO) Program Technology Provider Companies, comprising Hewlett Packard Enterprise, IBM Corporation, and Quantum Corporation, have released their annual tape media shipment report. Following sustained enterprise data expansion, the report highlights a strong start to 2026, driven by continued LTO-9 adoption and the rollout and initial capacity ramp-up of LTO-10 media.

According to the consortium, total shipped tape capacity in Q1 2026 increased 57 percent year over year compared to Q1 2025. This rebound follows an all-time record set in 2024, when total shipments reached 176.5 exabytes (a 15.4 percent annual increase). While total annual shipments settled back to 160.3 exabytes in 2025 (a 9 percent decline), the volume remained higher than any previous year on record prior to 2024. The arrival of the 40TB native capacity cartridge in early 2026 has provided an immediate density upgrade path for enterprise environments managing large-scale archival and secondary storage tiers.

Industry leadership notes that the ongoing increase in capacity requirements, power constraints in data centers, and the need for physically air-gapped cyber resilience are driving the integration of tape into modern hybrid architectures. The format is increasingly positioned as an economical cold tier to balance costly high-density flash and disk arrays, specifically for long-term retention and AI training datasets that require massive ingest capacity without constant active power draw.

Market analysis from industry observers indicates that, despite broader macroeconomic and international trade headwinds affecting enterprise procurement cycles, tape media maintains steady structural demand. The format’s low operational expenditure, zero idle power consumption per cartridge, and predictable media roadmap make it an effective hedge against data center power limitations. As organizations transition legacy libraries to current-generation hardware, the consortium expects total shipment volumes in 2026 to continue tracking toward new highs.

The post LTO Tape Shipments Up 57% in Q1 2026 as AI and Archive Demand Accelerate appeared first on StorageReview.com.

VDURA has completed the deployment of its data platform at New Mexico State University (NMSU), moving the system into full production. The infrastructure is designed to serve the university’s research community with a high-durability, high-throughput storage environment tailored specifically for artificial intelligence and high-performance computing (HPC) workloads.

NMSU, which holds Carnegie R1 status and manages over $141 million in annual research expenditures across aerospace, cybersecurity, agriculture, water science, and biomedical disciplines, has integrated the platform into its Research Cores Program. This program centralizes shared instrumentation and computational resources to accelerate large-scale simulation and AI model training across diverse academic departments.

Mixed-Fleet Architecture and InfiniBand Integration

Technically, the installation utilizes VDURA’s mixed-fleet storage architecture, integrating high-speed NVMe flash alongside high-density HDD storage tiers. This tiered configuration exposes a single global namespace across the university’s InfiniBand network fabric. The NVMe tier delivers low-latency responsiveness and high IOPS for active computational pipelines, while the high-density HDD tier provides cost-effective bulk capacity for secondary data sets and long-term retention.

Because the platform relies on a software-defined architecture supported by a subscription model, the university can independently scale performance and raw capacity over time to match evolving computational demands without disrupting active production workflows.

Alignment with Post-Quantum Cryptography Research

The production deployment builds on an existing strategic partnership established in August 2025 between VDURA and NMSU, focused on post-quantum cryptography (PQC). That initiative aims to develop and commercialize cryptographic methods to secure petabyte-scale data pipelines supporting AI and HPC workloads against emerging quantum computing threats.

By standardizing on VDURA hardware and software, NMSU runs its production research on the same storage infrastructure that underpins its ongoing security and data pipeline co-development projects. VDURA technical leadership noted that providing stable, high-efficiency data infrastructure maximizes compute-hour yield for high-stakes research environments, supporting the institution’s expanded computational requirements.

The post VDURA Deploys High-Performance Storage Platform for AI and HPC at New Mexico State University appeared first on StorageReview.com.

High availability has moved from a data center luxury to a line item in ordinary resiliency planning. Small businesses and edge sites now run point-of-sale, surveillance, and shared storage that cost money every minute they are down, yet the classic answer, a dual-controller enterprise array, is sized and priced for the data center rather than a branch closet or a wall-mount rack. That mismatch has left smaller IT shops with backups and snapshots for data protection but nothing for continuity: a single-controller NAS, however well-built, remains a single point of failure.

High availability is not new territory for QNAP, which has offered dual-controller hardware and replication-based recovery paths. What changes with QuTS hero h6.0 is efficiency. The new High Availability Manager builds clustering into the operating system on ordinary hardware, letting two identical NAS units form an active-passive cluster behind a single IP address, so that when one box goes down, the other takes over and clients barely notice. For an organization trying to right-size its IT investment, that is HA at the cost of a second NAS rather than a second class of infrastructure.

That is the pitch: to see how it holds up in practice, we built an HA cluster in the lab from two QNAP TS-h765eU systems loaded with Seagate IronWolf Pro hard drives and QNAP E1.S SSDs for cache, then set out to break it in the same way failures occur in edge locations. We pulled the power on the active node mid-transfer. We yanked its network connection while leaving the heartbeat intact. In both cases, the question was the same: does the file transfer survive, and what does recovery look like when the failed node comes back?

QNAP TS-h765eU Overview

The TS-h765eU is an interesting choice as an HA building block precisely because it’s relatively simple hardware. The enclosure is a 1U short-depth rackmount NAS, just 292.1mm (12 inches) deep, designed for small media cabinets, wall-mounted network racks, and edge deployments where a full-depth chassis will not fit. Each unit offers four 3.5-inch SATA drive bays up front, plus three E1.S/M.2 PCIe NVMe slots on the rear, giving it a hybrid storage personality: spinning disk for capacity, flash for caching or fast tiers.

QNAP has also designated it as a long-term supply model with availability guaranteed through 2031, which matters for organizations standardizing on a platform across sites.

Inside, the TS-h765eU runs Intel’s Atom x7405C, a quad-core chip that offers up to 3.4GHz, paired with 8GB of DDR5 memory, upgradable to 16GB, with In-Band ECC. Networking starts with dual 2.5GbE ports, with 10GbE available by swapping an E1.S bay for QNAP’s QXG-ES10G1T module.

Specification	QNAP TS-h765eU
CPU	Intel Atom x7405C quad-core, up to 3.4GHz
Memory	8GB DDR5, upgradable to 16GB (In-Band ECC)
Drive Bays	4 x 3.5-inch SATA
Flash Slots	3 x E1.S / M.2 PCIe NVMe
Networking	2 x 2.5GbE, 10GbE via optional QXG-ES10G1T E1.S module
Form Factor	1U short-depth rackmount, 292.1mm (12 in) deep
Operating System	QuTS hero h6.0 (ZFS-based)
Availability	Long-term supply model through 2031

 

For this evaluation, each node was populated with four Seagate IronWolf Pro 30TB hard drives (ST30000NT011) in a RAID 5 storage pool, plus two 3.84TB QNAP SSD700 E1.S drives (SSD700D1-003T84) in RAID 0 as cache. Both systems ran QuTS hero h6.0.0.3500 with High Availability Manager 2.0.421.

QuTS hero h6.0 and the HA Manager Architecture

High Availability Manager is the headline feature of QuTS hero h6.0, implementing a classic active-passive design. One NAS, the active node, serves all data and services. The second NAS, the passive node, continuously synchronizes with the active node over a dedicated heartbeat connection and stands ready to take over. Clients never talk to either node directly; instead, the cluster presents a single IP address and hostname, and whichever node is currently active responds to it. When a failover occurs, the cluster IP redirects traffic on the backend, so mapped drives, iSCSI initiators, and backup jobs do not need to be repointed.

The heartbeat link is the backbone of the design. QNAP requires a direct connection between the two units, with no switches in the path, and it carries both health-check traffic and block-level data synchronization between nodes. A separate cluster connection handles client-facing traffic through the normal network. Rounding out the split-brain protection is a quorum server, a third witness on the network that helps a node determine whether it should promote itself when the heartbeat goes quiet. We cover split-brain and the deployment topology that prevents it in detail later in this review.

QNAP says over 90 percent of NAS services are HA-ready in h6.0, and the cluster can extend capacity with JBOD expansion enclosures. There are caveats. Immutable snapshots, another marquee h6.0 feature, are not currently supported inside an HA cluster, so administrators will have to choose between the two protections for now. HA also requires two identical systems, matching model and firmware, which the pairing wizard enforces before allowing you to proceed.

Building the QNAP HA Cluster

Cluster creation starts from two independently configured TS-h765eU units with identical hardware. From the designated active node, the High Availability Manager wizard walks through a short pairing process. The wizard discovers the passive node over the heartbeat link, then asks you to assign roles to your network interfaces: a cluster connection that serves as the communication channel for access to the cluster, and the heartbeat connection, which QNAP notes is dedicated to syncing data and must be a direct link between the devices. In our build, the wizard assigned the heartbeat to Adapter 1 and the cluster connection to Adapter 2.

Next comes cluster identity. You assign a cluster hostname and a cluster IP address, which become the single address clients use from that point forward. Our cluster was named SR-Test at 176.16.248.34, sitting in front of nodes QNAP-HA1 (176.16.248.33) and QNAP-HA2 (176.16.254.157), the two nodes sitting in different /24 subnets on the lab network; the wizard paired them without complaint.

With settings confirmed, the wizard executes a five-step build: set up the heartbeat connection, set up the environment, stop services, configure system settings, and start services. The UI emphasizes that you should not power anything off during this window. On our systems, the build itself completed in about five and a half minutes, after which HA Manager reported the cluster created and redirected us to the cluster IP. Start to finish, going from two standalone NAS units to a serving cluster took under ten minutes of wizard time.

The heavier lift is the initial synchronization, in which the active node replicates its storage pool to the passive node, block by block. HA Manager surfaces this clearly with a progress bar, item count, and time estimate at the top of the dashboard, along with a warning that switchover and firmware updates are unavailable until sync completes. Watching the heartbeat statistics during this phase was one of the more impressive parts of the process: transfer speeds ranged from roughly 900 MB/s to 1.2 GB/s, with latency in the hundreds of microseconds, including one representative reading of 1 GB/s at 232 microseconds. Our initial sync completed in roughly ten minutes, right in line with the dashboard’s opening estimate of nine.

Once synchronization finishes, the dashboard settles into a Good status: cluster healthy, heartbeat connected, both nodes visible with live CPU, memory, disk throughput, and network statistics side by side, and per-pool synchronization status down below. It is a clean, information-dense view that answers the two questions an admin actually has: Is the cluster healthy, and is my data in sync?

Failover Test 1: Power Loss on the Active Node

Our first failure scenario is the blunt one: total power loss on the active node. We started a Windows file copy from a client to an SMB share on the cluster IP, a 41.1GB batch of six files including Windows 11 and Rocky Linux ISOs and a pair of Kali Linux VM archives, then pulled the power on QNAP-HA1 mid-transfer.

From the client’s perspective, the copy stalled for just under a minute, roughly 50 seconds, from the power pull to data moving again, and Explorer never surfaced an error; the transfer dialog held its progress, then picked back up as QNAP-HA2 promoted itself to active. HA Manager briefly reported the failover of the active node role in progress, then raised a warning banner: unable to detect passive node QNAP-HA1, with the suggestion to ensure the node is powered on and connected to the network. The transfer continued against the same cluster IP at full speed while the cluster ran on one node, which is exactly the degraded-but-operational state HA promises.

Restoring power to the QNAP-HA1 automatically triggered the reverse process. The node booted and rejoined as the passive member roughly ten minutes after losing power, and HA Manager began synchronizing data from the active node QNAP-HA2 back to QNAP-HA1, again with progress and time estimates visible on the dashboard, while our file copy continued undisturbed. Failback required no intervention: once the resync wrapped, the cluster paused the transfer for about 35 seconds while it returned the active role to QNAP-HA1, then let the copy run to completion. By the end of the run, the cluster was back to Good status with the transfer at 100 percent, having survived a hard power failure, a single-node period, a node restoration, and a failback within a single-copy job.

Failover Test 2: Network Loss with Heartbeat Intact

The second scenario is subtler and arguably more common in the real world: the active node loses its client-facing network connection (a failed switch port, a pulled cable, or a bad transceiver) while the node itself keeps running and the heartbeat link stays up. This is the case where split-brain protection is critical because both nodes are alive and can see each other, and the cluster must decide which node should own the cluster IP.

We repeated the same Windows file copy against the cluster IP and disconnected the cluster network interface on the active node. The transfer paused for about 45 seconds while the cluster moved services to the other node, then resumed without failing. HA Manager flagged the fault precisely, warning that cluster network interface Adapter 2 on the node was disconnected and suggesting a check of the switch connection, and then went a step further, noting that the disconnected node could no longer reach the quorum server through that interface. That specificity, naming the exact interface, node, and consequence, is more diagnostic than the generic degraded flag that some HA implementations settle for.

Reconnecting the network brought the node back into the cluster, and the failback came on its own within a minute: a second pause of roughly 30 seconds while the active role returned to QNAP-HA1 was the only client-visible evidence that anything had happened. The same copy job survived two switchovers in a row without a single failed file. For anyone who has watched an SMB session die from far less, that is the headline result of this entire exercise.

Deploying HA the Right Way: Split-Brain and Network Topology

Failover testing like ours proves the cluster works, but how well an HA pair holds up in production depends as much on the surrounding network as on the NAS units themselves. The scenarios HA Manager is designed for (a failed node, a dead switch port, a pulled uplink) all share one assumption: at least one communication path between the two nodes survives the failure. Protecting that assumption is the single most important decision in an HA deployment, because the alternative is the one condition every high-availability architecture is built to avoid: split-brain.

QNAP documents the scenario. Split-brain occurs when both nodes lose communication with each other but remain operational independently, each assuming the active role. Two nodes, each believing it owns the cluster, each willing to accept writes, are a recipe for data inconsistency or corrupted storage, since each may attempt to control shared resources simultaneously. The documented causes are exactly what you would expect: network disconnections between nodes, heartbeat failures, and unstable network paths. Note what these have in common. Split-brain is not triggered by a single node failure; it is triggered when all paths between two healthy nodes fail simultaneously. That is a topology problem, and it has a topology solution.

The deployment rules fall out as expected:

Run the heartbeat as a direct cable between the two nodes. QNAP requires a direct link for the heartbeat, and this is why. With no switch, no transceiver, and no shared infrastructure in the path, the only thing that can take the heartbeat down is a node itself, which is precisely the event failover is designed for. In a same-rack deployment, where these short-depth TS-h765eU units are likely to live side by side, there is no reason to do anything else.

If the nodes are separated, keep heartbeat and client traffic on physically independent paths. Where a direct cable is impractical, route the heartbeat through switches other than those used for the cluster connection. The moment both connections traverse the same switch, that switch becomes a single point of failure capable of severing all paths between the nodes simultaneously, turning an ordinary switch failure into a potential split-brain event. The whole point of buying two NAS units is defeated if one piece of shared infrastructure can isolate them from each other. This is the logic behind our network failover test methodology: pulling the client-facing lead while the heartbeat stays connected simulates the switch or cabling failure that a properly separated topology would actually experience, with the heartbeat alive to arbitrate the takeover cleanly.

Enable the quorum server. QNAP’s third safeguard is a witness on the network, configured under High Availability Manager > Settings > Failover Policy > Quorum Server. If the nodes lose their direct link but can still reach the network, the quorum server continues to monitor both and relays their status, giving each node an independent tiebreaker before it promotes itself. The quorum server’s connection status is displayed persistently on the HA Manager dashboard alongside the heartbeat, so its health is visible at a glance.

Should the worst happen anyway, QuTS hero handles split-brain defensively. Once connectivity is restored and the nodes can talk again, they exchange status information, recognize that both held the active role, and deliberately stop most services, including SMB and iSCSI, to avoid merging two diverged datasets. HA Manager then presents a Recover from Split-Brain wizard with two paths. Option one preserves data on a single node you select; the other node is wiped, reset as the passive member, and resynchronized, the fast route when you know which side has the correct data. Option two preserves data on both nodes by resuming services on one node while removing the other from the cluster entirely, allowing you to verify and reconcile the data before manually rejoining it. It is a sane recovery model, but recovery still means downtime and a full resync. Split-brain is a condition you prevent through deployment design, not one you plan to recover from, and the three rules above cost nothing beyond a cable and five minutes in a settings panel.

Monitoring and Management

Day-two operations live in the HA Manager app, which consolidates cluster status, per-node resource usage, event logs, and node management, including manual switchover, into a single pane. The dashboard’s warning banners proved specific enough to be diagnostic during our testing, naming the exact interface and node at fault rather than a generic degraded flag.

QNAP has also pushed HA awareness into the hardware itself. In an HA cluster, the LCD panel on each NAS can display the cluster name, the node’s current role, and the cluster IP, and the status LED indicates the HA state at a glance: solid green for the active node, blinking green for the passive node, and solid red for an HA error. In a rack full of identical short-depth units, being able to identify the active node without opening a browser is a small touch that techs will appreciate. For fleet deployments, AMIZcloud adds centralized cloud monitoring of HA groups, surfacing cluster health, latency, and alerts across sites.

Final Thoughts

High Availability Manager delivered on its core promise in both failure modes we threw at it. A hard power loss on the active node cost our Windows file copy roughly 50 seconds of stalled progress and zero failed files; a severed client network connection cost about 45 seconds. In each case, the same copy job then rode through node restoration and an automatic failback with nothing worse than a second sub-minute pause. Clients never had to repoint anything. Applications, file shares, and user workflows continued to operate through a single IP and hostname before, during, and after failover. Setup is approachable too; two standalone units became a serving cluster in under ten minutes of wizard time plus a ten-minute initial sync, and the dashboard told us exactly which interface, node, and connection to worry about every time we broke something.

The costs of HA shouldn’t be ignored, though. You are buying two of everything, and the passive node is idle capacity until the day it is not. Immutable snapshots, h6.0’s other marquee protection, are off the table inside an HA cluster today, so administrators must pick between the two. And the identical-hardware requirement means upgrades happen in pairs, which may constrain budgets.

Where this ends up is a question of right-sizing, and it is the small business and edge deployment that HA Manager serves best. Against dual-controller enterprise arrays, a pair of short-depth TS-h765eU units plays a different financial game entirely while covering the failure scenario, controller loss, that drives most dual-controller purchases in the SMB tier. Against DIY replication schemes, snapshot shipping, rsync jobs, and backup-and-restore, the difference is recovery time: those schemes protect data but leave you rebuilding client connections for hours, whereas HA Manager’s worst client-visible event in our testing was a minute-long pause. Just as important, the failure it cannot absorb, every inter-node path dying at once, is preventable with a direct heartbeat cable, separated network paths, and an enabled quorum server: a topology bill that totals one cable and five minutes in a settings panel.

One last point specific to a two-node design: the cluster’s most vulnerable window is a resync, when one node holds the only good copy of your data and the drives underneath are working their hardest. That is an argument for taking drive selection seriously in an HA pair, and it is why NAS-rated disks like the Seagate IronWolf Pro 30TB units in our build matter more here than in a standalone box: their job is to keep that window uneventful.

The QNAP TS-h765eU and QuTS hero h6.0 are available now. For more information, visit the QNAP TS-h765eU product page.

This report is sponsored by QNAP. All views and opinions expressed in this report are based on our unbiased view of the product(s) under consideration.

The post QNAP High Availability Tested: Sub-Minute Failover on a Pair of TS-h765eU appeared first on StorageReview.com.

Kioxia and Sandisk outlined a joint capital investment plan totaling over $31 billion (approximately 5 trillion yen) through 2032 to expand advanced 3D NAND flash memory production in Japan. The planned deployment of capital is contingent on Japanese government subsidies and builds on the joint venture’s 25-year manufacturing partnership, which has previously directed more than $50 billion into local fabrication infrastructure.

The capital expenditure will fund advanced tooling, AI-driven cleanroom automation, and infrastructure expansion across the joint venture’s primary manufacturing hubs at the Yokkaichi and Kitakami plants. Both organizations aim to secure multi-year bit growth and stabilize global flash supply as high-density storage demands accelerate across enterprise workloads, hyperscale infrastructure, edge deployments, and client AI architectures.

Fab3 Construction Underway at Kitakami Plant

In conjunction with the investment roadmap, Kioxia has initiated site preparation for Fab3, a new semiconductor fabrication facility on land south of Fab2 at the Kitakami Plant in Iwate Prefecture. Fab3 is designed to manufacture the next-generation BiCS FLASH 3D NAND technology, with initial production targeted for fiscal year 2029. Final deployment schedules, cleanroom buildouts, and equipment installations will be adjusted in phases based on market demand and finalized government support allocations.

This expansion follows the January agreement that formally extended the joint venture operating framework at the Yokkaichi facility through December 2034. The joint venture structure allows both companies to co-invest in wafer fabrication assets, share process technology development, and leverage smart manufacturing systems to scale production yields on leading-edge node transitions.

Leadership Perspectives on Long-Term Strategy

Executive leadership highlighted that sustained capital deployment remains critical to maintaining competitive bit density, power efficiency, and high-bandwidth output for demanding compute platforms.

“This joint investment further strengthens our longstanding partnership with Sandisk and underscores Kioxia’s strong commitment to contributing to the advancement of an AI-driven society,” said Hiroo Ota, President and CEO of Kioxia. Ota also pointed to strategic support from the Japanese government as essential to maintaining production competitiveness.

“For decades, Sandisk and Kioxia have jointly developed world-class NAND flash memory technology,” said David Goeckeler, Chairman and CEO of Sandisk. Goeckeler said the planned investments are in line with the company’s business strategy and financial guidance, ensuring supply for growing customer demand while creating new economic opportunities in the communities where the joint venture operates.

The post Kioxia and Sandisk Plan Over $31 Billion for Japan NAND Fabs Through 2032, With Kitakami Fab3 Targeting Fiscal 2029 appeared first on StorageReview.com.

IBM has announced the IBM Storage DS8000 Release 10.2, an update to its flagship enterprise storage platform tailored for IBM Z, IBM LinuxONE, and high-performance, mission-critical environments. The release targets large-scale enterprise demands by focusing on increased drive density, deeper operational telemetry, and automated cyber resilience.

Fifth-Generation FlashCore Module (FCM5)

Release 10.2 introduces the fifth-generation IBM FlashCore Module (FCM5), housed in the redesigned High-Performance Flash Enclosure (HPFE) Gen 3 E3. The architecture transitions the physical medium to the Enterprise and Datacenter Standard Form Factor (EDSFF), providing improved airflow, lower thermal overhead, and integration with Active Energy Management (AEM) for real-time power and cooling observability.

IBM says back-end storage fabric throughput and IOPS are doubled using NVMe over PCIe Gen 4 connectivity. This interface enhancement specifically benefits read-heavy, cache-unfriendly enterprise transaction workloads and large dataset processing.

FCM4 vs. FCM5

The shift from FCM4 to FCM5 represents significant architectural and density advancements across the FlashCore lineup.

Attribute	FCM4	FCM5
Introduced	2024	2026
Form factor	2.5-inch U.2 NVMe	EDSFF (Enterprise and Datacenter Standard Form Factor)
Enclosure	HPFE Gen 3 (prior design)	HPFE Gen 3 E3 (redesigned)
Interface	NVMe over PCIe Gen 4	NVMe over PCIe Gen 4
Max raw capacity	38.4 TB	Up to 76.8 TB
Max provisionable/effective capacity	Up to 115.2 TB (at 3:1 compression)	Up to 134 TB provisionable per drive
NAND technology	Charge Trap QLC NAND	Charge Trap QLC NAND (high-density)
Airflow/thermal design	Standard U.2 thermal profile	Improved airflow via EDSFF design
Power/thermal observability	Not a defined feature	Active Energy Management (AEM) with real-time power and thermal telemetry
Cyber resilience	Introduced in-drive AI ransomware threat detection	Extends in-drive AI threat detection to CKD data with near real-time alerting and automated CSM containment
Back-end throughput/IOPS	Baseline for prior generation	Doubled versus baseline, per DS8000 Release 10.2

Operational Visibility and Mainframe Integration

Release 10.2 integrates native IBM Storage Insights telemetry directly into DS8000 environments, consolidating health monitoring and performance metrics across both storage systems and SAN fabrics into a unified interface.

For mainframe deployments, the platform incorporates advanced zHyperLink performance statistics. These metrics provide storage administrators with granular monitoring of low-latency channel programs to maintain strict service-level agreements for core transactional systems.

Automated Cyber Resilience and Recovery

Expanding on the threat detection capabilities of Release 10.1, Release 10.2 pairs inline hardware AI with automated remediation workflows. The onboard module intelligence monitors continuous I/O streams for anomalous signatures and ransomware behavior on Count Key Data (CKD) volumes, issuing immediate alerts to operational teams and support channels.

Automated containment is driven via native Copy Services Manager (CSM) integration. When anomalous activity is detected, CSM triggers automated policy actions to isolate affected environments and reduce exposure windows. Furthermore, CSM introduces automated Type 1 infrastructure validation to verify that system environments restart cleanly from immutable Safeguarded Copies. Organizations requiring deep structural validation can extend these protections through IBM Z Cyber Vault and IBM Expert Labs Cyber Vault services.

The post IBM DS8000 Release 10.2 Brings FCM5 in EDSFF, Doubled Back-End Throughput, and Automated Ransomware Containment appeared first on StorageReview.com.

Gartner has published its 2026 Magic Quadrant for Enterprise Storage Platforms, and the headline is continuity at the top, with one clear standout. The same six vendors hold the Leaders quadrant as last year: Everpure, Huawei, HPE, NetApp, Dell Technologies, and IBM. Within that group, Everpure sits highest in ability to execute and furthest in completeness of vision for the second consecutive year, a position the chart makes unambiguous.

Six Leaders, Little Movement

The report, dated August 19 with positions as of July 2026, evaluates what is effectively the primary storage market; Gartner tracks the high-performance file and object crowd in separate research. Eight vendors appear in total. Behind Everpure, Huawei posts the second-highest ability to execute, HPE holds the strongest combined position of the remaining pack, NetApp and Dell cluster mid-quadrant, and IBM rounds out the Leaders. Hitachi Vantara stands alone in Visionaries; IEIT Systems is the lone Niche Player. When two-thirds of the evaluated field earns the Leader label, the label itself tells buyers little; the relative positions and Gartner’s cautions in the full report carry the useful signal.

What We Have Seen From the Leaders

Gartner’s placements track closely with what has crossed our bench and news desk over the past year. Everpure has had the busiest stretch, launching Data Stream for enterprise AI pipelines and expanding Data Intelligence and its Enterprise Data Cloud, while our lab put a FlashArray XL130 R5 at the center of an autonomous Fibre Channel SAN encryption project with Emulex SecureHBA.

Dell spent the year on the most aggressive enterprise storage reset in its lineup, PowerStore Gen 3, and recently pushed 9.83PB into 2U with 245TB SSDs on ObjectScale. HPE kept building out Alletra Storage MP, whose B10000 we took through a full multi-protocol deep dive, adding Kubernetes disaster recovery via CloudCasa and a broader AI factory portfolio expansion, while our own lab time with the X10000 and its Data Protection Accelerator Node showed what backup looks like without the bottleneck. More recently, we mapped a full-stack cyber resilience architecture on the B10000 against NIST CSF 2.0. NetApp extended its validated architecture play with FlexPod AI designs alongside Cisco, took sovereign, air-gapped deployments to Google Cloud, and remains the subject of our NetApp ASA deep dive on block-only ONTAP, which still holds up as the reference on that platform. IBM’s year has centered on the next-generation FlashSystem with agentic AI operations. Huawei remains a Leader on the strength of its OceanStor business outside North America, a market position Gartner continues to recognize even where our hands-on access is limited.

The full report is available from Gartner, with licensed reprints offered by several of the named vendors.

The post Gartner Magic Quadrant for Enterprise Storage 2026: Everpure Tops Both Axes Again as the Six Leaders Hold appeared first on StorageReview.com.

Dell has qualified KIOXIA’s 245.76TB NVMe SSDs for software-defined ObjectScale, putting 9.83PB of raw flash in a single 2U server. The density milestone lands on top of the data-path work Dell detailed earlier this year across ObjectScale 4.2 and 4.3, which targets high-throughput AI training, small-object pipeline efficiency, and inference acceleration through native S3 over RDMA, Key-Value (KV) Cache offload for LLMs, an overhauled chunk-store engine, and direct Apache Iceberg support via S3 Tables.

Density Milestone: 9.83 PB in a 2U Footprint via 245 TB KIOXIA SSDs

As enterprise AI data lakes expand across model training, fine-tuning, and compliance-retention tiers, rack density, thermal envelope management, and floor-space constraints have become primary design variables. To address raw capacity per rack unit, Dell has qualified KIOXIA’s 245.76 TB NVMe SSDs for software-defined ObjectScale running on Dell PowerEdge R7725 servers.

By populating 40 of the 245.76 TB drives in a single 2U chassis, ObjectScale reaches 9.83 PB of raw all-flash capacity; we reviewed the platform last summer. This density is aimed directly at enterprise AI factories, vector databases, large analytics repositories, and cyber-resilience platforms that require multi-petabyte unstructured storage footprints while minimizing power consumption, cooling load, and cabling complexity.

Throughput Gains and Small-Object Architecture

In software-defined deployments on qualified Dell PowerEdge servers, Dell’s internal testing indicates per-node read throughput of up to 40 GB/s, an 8x increase over previous-generation all-flash object platforms. Dell also points to a UK high-frequency trading customer sustaining roughly 280 GB/s of aggregate read throughput in production. Those figures line up with what we have measured on the hardware itself: in our R7725xd testing last November, the platform sustained more than 300 GB/s across the internal NVMe pool and 160 GB/s over NVMe-oF RDMA. That work characterized the server directly rather than running ObjectScale on top of it, but it establishes the ceiling the software is working against.

To eliminate metadata bottlenecks common in pipelines dealing with millions of small feature sets, vector chunks, and telemetry logs, ObjectScale uses an optimized chunk-store engine. The architecture aggregates small payloads into unified 128MB chunks before applying erasure coding across nodes. For typical 10KB files, more than 10,000 objects fit into a single chunk, substantially lowering metadata overhead.

This chunking approach also restructures drive failure rebuilds by reducing the total number of shards that must be recreated from billions to millions, cutting recovery times for dense NVMe drives from weeks to hours. Inline checksums verified at the stripe level reduce host CPU cycles spent on background integrity scans. Furthermore, the re-engineered Key-Value Store in ObjectScale 4.2 provides roughly 4x better memory efficiency and cuts metadata disk footprints by 30% to 60%.

S3 over RDMA and Inference Acceleration

ObjectScale 4.3 expands S3 over RDMA across Dell’s all-flash portfolio, including the XF960, EXF900, and software-defined PowerEdge R7725xd deployments. Utilizing RoCEv2 networking and Dell’s S3-over-RDMA SDK, the platform bypasses kernel TCP overhead and CPU interrupts to establish direct memory paths between host GPUs and backend NVMe media. Internal testing shows this delivers up to 230% higher throughput, an 80% reduction in latency, and up to a 98% reduction in client CPU utilization compared to standard S3 over TCP.

For inference architectures, ObjectScale functions as an external tier for KV Cache offload. Because long-context inference quickly exhausts on-GPU HBM and VRAM, integrating with vLLM, LMCache, and NVIDIA’s NIXL library allows inference frameworks to spill active attention states directly to ObjectScale over RDMA-accelerated S3. Dell reports up to a 19x improvement in Time to First Token against standard vLLM without cache offload, along with up to 5.3x higher token throughput and nearly 3x higher multi-turn throughput. Separately, Dell cites a TTFT of roughly 0.86 seconds in its own head-to-head testing against a competing object platform.

Native S3 Tables and Scale-Out Efficiency

ObjectScale 4.3 also brings native S3 Tables, which Dell lists as a tech preview, embedding Apache Iceberg open table formats directly within the storage bucket layer. Query engines including Trino, Starburst, Spark, and Flink can run analytics natively against S3 buckets without requiring dedicated staging or ETL data movement into separate warehouses. Dell reports that this architecture delivers up to 2x faster ingestion and up to 4.5x faster queries compared to traditional warehouse configurations.

Underlying storage efficiency includes 24+2 and 24+4 erasure coding profiles, which Dell says cut write amplification by up to 75%. On high-capacity HDD platforms like the EX500, the company reports up to 25% faster large-object ingest and up to 2x higher mid-size object write performance. Workload-specific bucket compression supports Snappy, LZ4, ZSTD, and Deflate algorithms with telemetry for FinOps cost tracking. Provisioning and orchestration integrate natively with cloud-native tooling, including Kubernetes Container Object Storage Interface (COSI) and Terraform providers.

The post Dell Puts 9.83PB in 2U With 245TB KIOXIA SSDs, Adds S3 over RDMA and KV Cache Offload to ObjectScale appeared first on StorageReview.com.

The Enterprise NAS (ENAS) is Ubiquiti’s first run at more enterprise-lite business storage. The 3U chassis holds 16 hot-swap drive bays, an eight-core Arm Neoverse N2 processor, 64GB of ECC memory, dual 25GbE SFP28 ports, and redundant hot-swap 550W power supplies, all for $3,999 MSRP. One pricing note: Ubiquiti’s store currently applies a memory surcharge at checkout, its response to the memory price environment, so out-the-door pricing runs above list until component costs settle. The filesystem underneath is ZFS, the drive bays accept whatever disks you want to put in them, and every software feature ships with the box. There are no licenses, no feature unlocks, and no mandatory support contracts.

The price for a bare unit might look daunting at first, but there is a lot of value to be found. Dual 25GbE as standard equipment is the piece competitors do not match: Synology’s 16-bay RS4021xs+ ships with two 10GbE RJ45 ports and puts 25GbE behind a PCIe add-in card, and the story is much the same across rackmounts at this capacity point, where faster-than-10GbE networking is another line on the quote. Pair that with redundant power and ECC memory, and Ubiquiti is checking boxes that have traditionally separated SMB NAS boxes from entry-level enterprise arrays. ENAS is available now directly from Ubiquiti (affiliate link).

ENAS also caps a portfolio expansion that has been running at a remarkable pace. Ubiquiti built its name on access points and gateways, and the networking side keeps rolling; in the past year alone, we have looked at the UniFi E7 and E7 Campus WiFi 7 access points, the Cloud Gateway Fiber, and the Dream Router 7. Storage is newer ground. The line started with the UNAS Pro, a $499 seven-bay 2U unit that made sense mostly for UniFi loyalists, then grew with the compact UNAS 2 and the UNAS Pro 8, which added redundant power and NVMe caching. ENAS is the first of the family with ZFS, ECC memory, native iSCSI, and SAS expansion ports, and it is worth noting that the file systems differ enough that Ubiquiti says UNAS data cannot be restored directly to an ENAS; only users, groups, and settings carry over.

The segment ENAS enters is not empty. Synology and QNAP, along with others, have owned the SMB NAS market for two decades, and their platforms remain far deeper in software: app ecosystems, containers, surveillance suites, and mature backup tooling that UniFi Drive does not attempt to match today. But the incumbents have provided challengers with an opening. Synology spent most of 2025 enforcing a compatibility policy that effectively locked its 2025 Plus-series units to validated, largely Synology-branded drives, before walking the policy back with DSM 7.3 in the fall. Against that backdrop, a 16-bay ZFS system with an explicit open-drive stance and no software licensing is a well-aimed product.

The buyer this makes the most sense for is easy to picture: a small or mid-sized shop, or the MSP that runs it, with a moderate-to-large UniFi estate already in place. For those environments, ENAS drops into the same console as the switches, gateways, access points, and cameras, inherits Site Manager for multi-site visibility, and ties into the same identity model, with UniFi Endpoint handling remote file access. Storage stops being a separate vendor with a separate UI and a separate renewal. For shops that need easy, reliable file and block storage and already understand the UniFi interface, operational continuity is as much the product as the hardware.

One note on our test unit: we have had ENAS in the lab since November 2025, and the production hardware revision moved the rear I/O from 10GbE to the dual 25GbE configuration shipping today. Everything pictured and tested here reflects the current retail hardware, and the software has been moving quickly since our unit arrived, a point we will come back to.

Ubiquiti Enterprise NAS Specifications

Specification	Ubiquiti Enterprise NAS
Overview
Dimensions	481.4 × 480 × 132 mm (19 × 18.9 × 5.2 in)
Storage Capacity	16 × 2.5/3.5-inch drive bays 2 × M.2 NVMe bays
Networking Interface	2 × 25G SFP28 (25G/10G/1G) 1 × 10GbE RJ45 (10G/5G/2.5G/1G/100M)
Expansion Port	2 × SFF-8644 (24G)
Power Redundancy	Supported
Form Factor	3U rackmount
Hardware
Drive Support	16 × 2.5/3.5-inch HDD/SSD 2 × M.2 NVMe SSD 2 × Expansion ports
Max Power Budget for Drives	450W
Max Power Consumption	550W
Power Method	Dual AC input, hot-swappable power modules
Power Supply	2 × Hot-swappable AC/DC 550W power modules
Processor	Eight-Core ARM N2 at 2.4GHz
Memory	64GB
Management	Ethernet
RF Interface	Bluetooth 4.1
Weight	16.1 kg (35.5 lb)
Enclosure Material	SGCC steel
Mount Material	SGCC steel
Supported Rack Depth	Rails support 600 mm (23.6 in) four-post racks with square holes (9.5 × 9.5 mm) Post depths from 600–1066 mm (23.6–42 in)
Faceplate	3U Bezel (4.7-inch touchscreen, RGBW LEDs) 3U Bezel Lite (blank) Both optional
LEDs
Status LEDs	Ethernet, SFP28, HDD, System, Expansion Port, CRPS
Environmental & Compliance
Operating Temperature	-5°C to 40°C (23°F to 104°F)
Operating Humidity	5% to 95% noncondensing
NDAA Compliant	Yes
Certifications	FCC, CE, IC
Software
Supported File Protocols	NFS, SMB
Supported Block Protocols	iSCSI
RAID Types	Mirror, RAID-Z1, RAID-Z2, RAID-Z3
RAID Groups	Multiple
Hot Spare Support	Supported
Personal & Shared Drives	Supported
SSD Cache	Supported
Max NVMe SSD Capacity	8 TiB
File Encryption	Supported
Backup Support	Remote UNAS, CIFS/SMB server, cloud services
Cloud Backup Services	Google Drive, OneDrive, Dropbox, Amazon S3, Backblaze B2, Wasabi
Snapshots	Supported
Share Links	Supported
Time Machine Backup	Supported
Client App Support	Supported
User Groups	Supported

Build And Design

The ENAS is built to a different standard than the rest of the UniFi storage line, closer to the rackmount servers that pass through the lab than to the UNAS Pro it descends from. The differences start at the drive bays. ENAS caddies use a proper release button with the activity LED set into its center, replacing the push-to-release caddies on the UNAS Pro. It is a small change that makes hot-swaps less fiddly when the rack is dark and a drive needs pulling. A plug at the top right of the front panel covers a connector we will come back to at the end of this section.

The aluminum handles on the rack ears do two jobs: they release the rackmount latches, and they anchor the optional bezel. Ubiquiti ships the ENAS with King Slide rails, the same family we see on servers from the major OEMs, and mounting them into the rack is tool-less. Getting the inner rail onto the chassis is not easy and takes a screwdriver and the screws along the bottom of the unit. Once that is done, the ENAS slides in and locks with the latches at the top of each handle.

The caddies are also where some cost engineering shows. The release lever is plastic, whereas the UNAS Pro used what feels like aluminum, a downgrade on the one part that gets handled every time a drive moves. Mounting is tool-less for 3.5-inch drives, with an optional locking screw for anyone who wants it, but 2.5-inch drives still take four screws through the bottom of the caddy. Fill all 16 bays with SSDs, and that is 64 screws standing between you and a populated chassis.

The rear panel carries an RJ45 console port, twin 25G SFP28 ports, and twin SFF-8644 connectors, better known as external Mini-SAS HD. Each Mini-SAS port carries 24G to a shelf, supporting up to two of the UACC-ESE-16 Enterprise Storage Expansion, a no-frills 16-bay expansion chassis. Ubiquiti specifies the shelf as a 3U unit with a single SFF-8644 port and the same pair of hot-swap 550W power modules used in the ENAS, and lists it as compatible with the ENVR-Core as well. It appeared on Ubiquiti’s site while we were finishing this review but had not reached the store, so we have not had one in the lab. Two shelves take the system to 48 bays, which is where Ubiquiti’s claim of more than a petabyte of raw capacity comes from.

A blank panel to the right of the expansion ports opens onto two M.2 slots. Trays are not included, so populating them means adding Ubiquiti’s UACC-SSD-Tray (Affiliate Link). UniFi Drive uses these as a read-only SSD cache rather than as a general-purpose fast tier. Ubiquiti does offer a read-write cache mode on the UNAS line, but the ZFS-based ENAS does not support it.

Further to the right sit the redundant hot-swap power supplies, a pair of 550W CRPS modules. Nothing about them is novel to anyone who works on servers, which is rather the point: this is the first UniFi storage box whose parts bin looks like the rest of the rack. Each module releases with a locking tab on the side and pulls out on a foldaway handle.

For a $3,999 chassis, the back panel is dense. Console, dual 25G, and SAS expansion on one plate is a spec sheet normally found a tier or two further up, and it is the clearest signal that ENAS is not simply a bigger UNAS.

And finally, the small port mentioned on the front of the unit at the beginning of this section is a magnetic USB Type-C connector for the optional UniFi 3U Bezel, which Ubiquiti specs with a 4.7-inch touchscreen and RGBW status lighting drawing under 4W. If screens and lights are not your thing, but you want to cover your drive bays, Ubiquiti offers the blank UACC-3U-Bezel-Lite. Both bezels are listed on Ubiquiti’s site but have not reached the store as of this writing, so we have not had either one in hand.

That bezel is more than decoration, because the ENAS has no power button. Shutting the system down from the interface is easy; getting it back up is not. With no front-panel control, the only way to power the unit on while standing at the rack is to pull both power cords and reseat them. Remotely, there is no power-on control in the UniFi interface either, which leaves Wake-on-LAN as the only option. For a box destined for closets and small racks that may never see a KVM, that is worth knowing before it ships to a site an hour away.

The build lands where Ubiquiti hardware usually does, clean on the outside with most of the practical details right: tool-less rail mounting, tool-less 3.5-inch caddies, redundant power, and 25G on board. The Mini-SAS ports are the part that changes the math later, since they turn a 16-bay box into a 48-bay one without replacing anything already installed. The missing power button is the one real misstep.

ENAS Management

ENAS is managed through UniFi Drive on UniFi OS, which means anyone who has adopted a UniFi device in the past will find nothing surprising in the workflow. The system appears in the same console as the rest of a UniFi deployment, and Site Manager provides fleet-level visibility across sites so that an MSP can watch capacity and health on a client’s ENAS alongside their switches and access points. Local storage administration covers ZFS pool creation in Mirror, RAID-Z1, RAID-Z2, or RAID-Z3 layouts, shared folders over SMB (including Time Machine targets for macOS clients) and NFS, and iSCSI LUNs for block workloads. Scheduled snapshots, replication between UniFi storage devices, and backup jobs to third-party cloud targets, including Amazon S3, Backblaze B2, Wasabi, Google Drive, OneDrive, and Dropbox, round out the data protection story, and AD/LDAP integration handles user import for shops with existing directories. Drive encryption is supported, as are ZFS inline compression and deduplication, though Ubiquiti sensibly cautions that dedup is resource-hungry and best reserved for SSD-backed pools.

The two external M.2 slots serve as an SSD cache rather than a general storage tier. Ubiquiti offers read-only and read-write cache on the UNAS line, although for the ENAS, the SSD cache is read-only with one or two SSDs. The cache deliberately bypasses large sequential files, since streaming them through the SSDs would burn endurance for no real gain, so the payoff lands in concurrent random access, office file sharing, and database-style workloads rather than large media pulls. We tested with and without the cache populated, and the results below bear that behavior out.

The caveat is that the feature set is still thin next to DSM or QTS. There is no application ecosystem, no container or VM hosting, and no equivalent of Synology’s Active Backup suite; ENAS is a storage appliance, not an app platform yet. The counterweight is how fast the platform is moving. Since our unit arrived, Ubiquiti has shipped a steady cadence of UniFi Drive releases, most recently version 4.3.10 on July 30, and the company has publicly committed to immutable snapshots, multi-site backup orchestration, and expansion shelves that Ubiquiti says will take a single system past a petabyte of raw capacity through the two SFF-8644 ports. Buyers should evaluate ENAS on what it does today, but the trajectory is important to consider, and the gap is closing quarter by quarter.

The landing page in UniFi Drive condenses the entire system into a single view. A status rail on the left covers the chassis: network address, uptime, firmware level (UniFi OS 5.1.19 and Drive 4.3.6 during our time with the unit), CPU temperature, and a three-position fan profile slider. The center pane tracks live disk throughput, while the panel below breaks out the pool with per-share usage and a snapshot schedule column, each share a click away from protection. Our RAID-Z2 pool showed 132.76TB in use of 170.57TB usable, with dedupe and iSCSI carve-outs itemized on the capacity bar.

Storage Pools are where the ZFS decisions get made, and Ubiquiti does more hand-holding here than most. Illustrated cards across the top explain the trade-offs between a single RAID group, multiple RAID groups in one pool, multiple pools with different protection levels, and global hot spares, useful guardrails for an admin who has never touched ZFS. Below that sits our test pool: a single RAID-Z2 group of eight 30TB 7,200RPM hard drives, good for two-drive failure protection, with bays 9 through 16 left open.

Block storage gets its own view. LUN creation is a short wizard, and the list shows host count, map status, per-LUN usage, and the backing pool at a glance. Initiator access is controlled through host groups rather than per-LUN fiddling; we provisioned a pair of 25GB LUNs mapped through a single host group for our iSCSI testing. It is minimal compared to what a dedicated SAN interface exposes, but it covers the Proxmox/VMware/Hyper-V shared-storage case that ENAS is aimed at.

The Services page collects protocol plumbing in one place: the web access endpoint, SMB with copy-ready paths for macOS and Windows, NFS exports with per-host permissions and squash settings alongside a pre-built mount command, plus Time Machine and Rsync toggles. One quirk worth noting is that file services use their own credentials, separate from UniFi accounts; a banner reminds you that they are required before SMB mounts or Time Machine will work.

Day-to-day administration happens in the file view, where shared drives are listed with active user counts, usage, pool assignment, and protection status. This is where our 314T share, holding the bulk of the 125TB test data set, lived alongside a pair of smaller shares. Personal drives are provisioned per user through Admins & Users, which keeps end-user storage separate from departmental shares without extra configuration.

Performance Testing

To exercise the ENAS over its shipping 25GbE interfaces, we built a single RAID-Z2 group from eight Seagate IronWolf Pro 30TB CMR drives in bays 1 through 8, leaving the back half of the chassis empty, and generated load from a Dell PowerEdge R740 running Ubuntu Server with a dual-port NVIDIA ConnectX-4 25GbE NIC, directly attached to the ENAS SFP28 ports. Our testing used FIO sweeps of 4K random and 1M sequential transfers, in both read and write, across a matrix of queue depths and job counts to find each protocol’s peak and its saturation behavior.

We ran the sweep against six configurations: NFS with asynchronous mounts, NFS with synchronous mounts, and iSCSI, each with and without the M.2 cache populated. The cache drives are Ubiquiti-supplied 1TB M.2 units that report as Kingston OM8SEP41024Q-A0, a TLC NVMe part from Kingston’s OEM catalog rather than an enterprise SSD. For a read-only cache that is a sensible fit, since the device serves reads and never has to absorb sustained writes. Charts below plot all six against each other; peak figures are called out in the text.

4K Rand Read IOPS

Random 4K reads cluster tightly across every NFS configuration, peaking between 89.5K and 90.9K IOPS regardless of sync mode or cache. With a working set this size against 64GB of ECC memory, ZFS is serving the bulk of these hits from ARC, so the ceiling reflects the network path and protocol stack more than the spinning disks underneath. iSCSI tops out lower, at 40.4K IOPS (39.7K with the cache populated), a little under half the NFS figure.

4K Rand Read Latency

Latency tells the same story from the other side. At low queue depths, every configuration answers 4K reads in 0.13 to 0.15ms, which is memory-and-wire territory, not disk. Response times scale predictably as the sweep deepens, and the NFS configurations stay well-behaved through moderate depths before climbing steeply once the box saturates at the extreme corners of the matrix. The practical read: ENAS has plenty of headroom for the concurrent small-file access an office generates.

4K Rand Write IOPS

Random 4K writes separate the configurations dramatically. Async NFS peaks at 9.6K IOPS (10.1K with cache), respectable for a RAID-Z2 pool of spinning disks. Force synchronous writes, though, and throughput collapses to a few hundred IOPS, with mean latencies in the tens of milliseconds; every commit is waiting on the pool. UniFi Drive does not expose a dedicated ZFS write-log (SLOG) device, which is the usual fix for exactly this behavior, and the ENAS SSD cache is read-only, so there is nothing in the platform today to absorb synchronous writes. iSCSI landed at 4.5K IOPS in the base configuration and 15.2K with the cache populated, an unexpectedly large spread for a read cache, so treat the delta between those two runs with some caution.

4K Rand Write Latency

The write-latency chart mirrors the IOPS results: async NFS holds sub-millisecond response at the low end of the sweep and degrades gracefully, while the sync configurations start near 20ms and climb from there. Nothing here is surprising for ZFS on hard drives without a log device. Still, shops planning sync-write-heavy workloads, databases, or NFS datastores with sync enforced should look at the async and iSCSI numbers and plan accordingly.

1M Seq Read GB/s

Large sequential reads are the ENAS headline result. Async NFS peaked at 4,170 MB/s, and all four NFS configurations held above 4,100 MB/s, comfortably beyond what a single 25GbE link can carry, confirming the system can drive both SFP28 ports at once. For a $3,999 box full of hard drives, north of 4GB/s of sustained read throughput is a legitimately strong showing that covers backup verification, media pulls, and multi-client streaming with room to spare. iSCSI sequential reads are the sore spot: the target peaked at just 257MB/s (227MB/s with cache) and degraded as load increased, behavior consistent with the target oversaturating rather than throttling cleanly. This looks like the kind of issue a firmware release resolves; until then, iSCSI on ENAS is better suited to random-access workloads than large sequential streams.

1M Seq Read Latency

Sequential read latency stays orderly on NFS, opening around 1.7ms and scaling with queue depth as expected. The iSCSI curve is the outlier, with response times an order of magnitude higher at equivalent load, which is the same oversaturation signature seen in the throughput chart.

1M Seq Write GB/s

Sequential writes over async NFS reached 2,567MB/s (2,654MB/s with cache), roughly the ceiling we would expect the RAID-Z2 pool itself to sustain. The sync NFS configurations tell the same story as the 4K results, topping out at 137 to 142MB/s. iSCSI writes fared far better than iSCSI reads at 1,432MB/s, further evidence that the read-path anomaly is a software issue rather than a hardware limit.

1M Seq Write Latency

Write latency closes out the sweep with no surprises: async NFS holds low single-digit milliseconds at sane depths, sync NFS lives in the tens of milliseconds, and iSCSI sits in between. Taken together, the data says ENAS is an excellent sequential machine and a capable async random performer, with sync-write workloads being the one area that demands planning.

Conclusion

The hardware value here is real. For $3,999, ENAS delivers 16 open bays, ZFS with ECC memory, dual 25GbE on board, redundant power, and expansion ports that will eventually take it past a petabyte, a combination the incumbents do not match at this price without add-on cards and, in some cases, drive-compatibility strings attached. Our testing backs the core proposition: over 4.1GB/s of sequential NFS read throughput, 2.6GB/s writes, and roughly 90K IOPS of cached 4K random reads is more performance than most small businesses will ever pull through it.

The decision comes down to software. Shops that live in Synology or QNAP’s application ecosystems, containers, surveillance, turnkey backup suites, will find UniFi Drive spartan by comparison, and virtualization builders should weigh the iSCSI sequential-read anomaly and the sync-write penalty before pointing a datastore at it. But that is not really who this box is for. For the SMB with a moderate or large UniFi estate that needs easy, reliable file and block storage, managed from the console it already knows, with snapshots, replication, cloud backup, and directory integration included and no license invoice ever arriving, ENAS is the most credible storage product Ubiquiti has shipped, and the first one the incumbents need to take seriously. Given the pace of UniFi Drive releases since our unit arrived, we expect the software gap to keep narrowing, and we will revisit the platform as immutable snapshots and the expansion shelves land.

Product Page – Ubiquiti ENAS

Ubiquiti Enterprise NAS (ENAS) (affiliate link)

The post Ubiquiti Enterprise NAS Review: 16 Bays, Dual 25GbE, and ZFS for $3,999 appeared first on StorageReview.com.

WEKA announced that Andromeda, the platform that connects AI teams with high-performance compute across an open market of providers, is integrating WEKA NeuralMesh as a core data storage layer for its managed GPU clusters. Andromeda customers can choose between a dedicated NeuralMesh deployment and a GPU-native NeuralMesh Axon configuration, giving them consistent storage performance on any cluster running at any hyperscaler or AI cloud.

Andromeda operates across more than 50 compute providers worldwide, routing training and inference workloads through a growing fleet of managed clusters, and says it serves more than 100 AI customers, supporting billions of GPU-hours. Every cluster on the platform must meet the same quality benchmarks, regardless of who operates the infrastructure, and Andromeda certifies each one across GPU, storage, network fabric, and security before it reaches a customer. That standard exposed the problem: each provider brought its own storage environment, and performance varied from cluster to cluster.

NeuralMesh aims to standardize that data layer. For Andromeda, the platform provides a deployable storage architecture that is consistently configured across hardware providers, rather than relying on each cloud or data center’s native shared-storage implementation.

“Our goal is to enable the global flow of compute, and that means every cluster we deliver has to perform, no matter where the capacity comes from,” said Wil Moushey, CEO at Andromeda. “Standardizing on WEKA NeuralMesh has given our customers hyperscaler-grade consistency with open-market flexibility at the storage and memory layer, where performance is won or lost. Idle GPUs are throttling the pace of AI innovation. WEKA helps us ensure every GPU we manage is earning its keep.”

GPU-Native Storage With NeuralMesh Axon

NeuralMesh Axon is the GPU-native deployment of the NeuralMesh software platform. It fuses storage directly into Andromeda’s GPU servers, converting each cluster’s existing NVMe into a unified data layer that feeds training and inference. The economic argument follows from that: the hardware is already racked, powered, and paid for, so the performance arrives without adding footprint, power draw, or cost. WEKA cites one case where a research lab found its workloads bottlenecked by the storage provided by its cluster, and Andromeda deployed Axon on the same hardware rather than adding infrastructure.

Andromeda uses the NeuralMesh Kubernetes Operator to automate cluster deployment and lifecycle management. The company reports that new NeuralMesh Axon clusters can be deployed from bare metal in about 10 minutes. Roughly half of its NeuralMesh deployments use the Axon configuration, while the rest are dedicated deployments for customers requiring exclusive GPU compute and memory resources.

“Andromeda’s first WEKA NeuralMesh Axon deployment took just 10 minutes to stand up from bare metal. That became our blueprint,” said Vishvajit Kher, lead architect at Andromeda. “If a cloud provider doesn’t offer shared storage, we’re no longer blocked. We deploy WEKA NeuralMesh and know it will perform. It’s a full-featured system, up to 90% less expensive than market alternatives, delivering savings that compound as we scale.”

The deployment model enables Andromeda to bring up managed GPU clusters in environments where cloud providers do not offer shared storage. This broadens the range of provider infrastructure that meets its operational requirements.

Reported Production Results

Andromeda reports that NeuralMesh Axon clusters sustain more than 400 GB/s of aggregate throughput per cluster and exceed 6 million IOPS in production deployments. In one biotech workload involving metadata-intensive datasets with about one billion files per directory, the platform reduced data transfer time from hours to minutes.

The company also reports that AI teams using its managed clusters reduced environment startup times from about three minutes to 30 seconds. Faster environment initialization can increase the number of training, testing, and inference iterations teams can execute during a development cycle.

NeuralMesh includes fault-isolation capabilities to prevent localized failures from cascading across a cluster. For providers operating multi-tenant GPU environments, this helps maintain workload availability while reducing the operational impact of infrastructure faults.

Extending Across a Global AI Compute Network

The WEKA deployment is expanding across Andromeda’s global infrastructure, including dedicated storage configurations and GPU-native Axon systems. The companies also plan to extend the integration to newer generations of AI infrastructure as Andromeda adds capacity and provider coverage.

For managed GPU providers, the deployment highlights the growing need for a portable AI data platform. GPU capacity alone does not guarantee predictable AI performance when storage and metadata behavior vary across providers. Standardizing the storage layer helps maintain throughput, IOPS, and operational workflows across a distributed fleet of heterogeneous GPU clusters.

The post Andromeda Standardizes Storage Across 50+ GPU Providers on WEKA NeuralMesh appeared first on StorageReview.com.

NetApp has acquired DataPelago, a California-based AI data infrastructure company. The acquisition extends NetApp’s data infrastructure portfolio with GPU-accelerated processing capabilities that operate at the storage layer, reducing the need to move enterprise data into separate compute environments before it can be used for AI and analytics workloads.

DataPelago’s core technology, Nucleus, is a universal data processing engine that leverages diverse CPU and GPU resources directly on data in place. Instead of transferring data from operational systems to separate analytics or AI clusters, Nucleus employs software-defined acceleration, keeping the data where it is stored. This zero-copy method addresses a key challenge in enterprise AI deployment by reducing costs, latency, governance issues, and infrastructure overhead caused by data movement.

DataPelago reports that Nucleus can cut infrastructure costs by as much as 80% and boost processing speed up to 10 times, compared to traditional architectures that separate storage and compute. These benefits vary based on workload and deployment setup, but the core strategy reflects a wider industry trend to bring computing closer to the data.

The technology allows NetApp to integrate data preparation and processing directly into its storage and data management platform. The company sees the acquisition as a way to help enterprises activate governed data for AI, improve GPU resource use, and enhance AI models. Fragmented data environments can lead to expensive accelerator infrastructure sitting idle while waiting for data pipelines, copies, transformations, and transfers instead of running model training or inference.

Following the acquisition, DataPelago operates as a wholly owned subsidiary of NetApp. Its engineering team and Nucleus technology are expected to bolster NetApp’s strategy in both on-premises and cloud data environments, where customers increasingly demand integrated data management and high-performance AI infrastructure.

The acquisition continues NetApp’s strategy of ecosystem partnerships with Cisco, Google Cloud, Red Hat, and SK Telecom. Bringing DataPelago into the fold enables NetApp to have a more direct role in AI data processing, especially for organizations aiming to minimize data transfers between storage systems, GPU clusters, and AI platforms.

The post NetApp Buys DataPelago to Run GPU Data Processing Where the Data Already Lives appeared first on StorageReview.com.

WEKA has launched two connected products: NeuralMesh 6, its most significant software release to date, and WEKApod 3, a new generation of storage appliances designed and engineered by WEKA to run it. NeuralMesh 6 remains deployable on customer-selected hardware, while WEKApod provides the turnkey path, shipping with the software preinstalled. NeuralMesh 6 introduces native multi-tenancy, a combined file-and-object protocol stack, metadata-driven data mobility, always-on data reduction with contractual guarantees, Kubernetes-native operations, and integrated observability. WEKApod 3 is custom-designed hardware built to run this software at what the company claims is the highest capacity and performance density available in a single rack, with three configurations, Nitro, Prime, and Prime Max, targeted at maximum performance, balanced capacity, and maximum density, respectively.

Both announcements frame the release around a common market shift: as AI workloads move from training toward long-context, agentic, and retrieval-driven inference at production scale, storage and memory infrastructure, not just GPU count, increasingly determine cost per token and achievable throughput.

NeuralMesh 6: Multi-Tenancy, Unified Protocols, and Data Mobility

NeuralMesh 6 combines features that WEKA claims AI infrastructure operators have traditionally needed to piece together from various vendors: multi-tenancy, a unified protocol stack for files and objects, data mobility across sites, continuous data reduction, Kubernetes-native management, and integrated observability, all integrated into a single software stack instead of assembled from separate parts.

Multi-tenancy is split into two tiers that can be combined. Composable Clusters provide hardware-level isolation, with dedicated CPU, memory, and storage drives per tenant, intended for anchor tenants that need guaranteed resources and predictable performance. Virtual Multi-Tenancy adds VPC-style network isolation through WEKA’s Virtualized RDMA Data Fabric, supporting private VLANs, overlapping IP address spaces, per-tenant quality of service and encryption with independent key management, and independent LDAP or Active Directory authentication per tenant. Virtual Multi-Tenancy scales to more than 1,000 isolated logical tenants per cluster, with new tenant provisioning in under 30 minutes. The two tiers compose, so a single WEKA hardware cluster running 50 Composable Clusters can support up to 50,000 logically isolated tenants on the same physical infrastructure, enabling growth from dozens to tens of thousands without re-architecting the system.

On the protocol side, NeuralMesh 6 implements a native S3 stack in which the same physical data blocks are addressable via S3 and POSIX simultaneously, rather than via a gateway that translates between them. A file written via NFS or POSIX is immediately readable from S3, and vice versa, eliminating duplicate full-dataset copies that typically accumulate as data moves between training, fine-tuning, and inference stages. WEKA built the S3 implementation specifically for AI access patterns, supporting 2,000 to 5,000 concurrent S3 connections per node, roughly five times the concurrency of conventional S3 architectures, with S3 over RDMA enabling zero-copy transfer directly into GPU memory.

Data mobility is handled through metadata-first replication, making a destination environment immediately browsable rather than requiring a complete data copy before a workload can start. Data hydrates on demand, reducing WAN traffic and allowing organizations to place workloads wherever GPU capacity exists rather than where the data was originally written. This release adds asynchronous replication and remote caching as a first step toward broader federation and a global namespace across sites and clouds.

That replication capability is already underpinning real deployments: Sam Tabar, CEO of WhiteFiber, said NeuralMesh’s intelligent replication lets the company make datasets visible across sites and pull exactly the data each job needs to the next GPU allocation, the same architecture behind Project Redwood, the 111.2 Tbps cross-data-center supercluster we covered earlier this month.

Data Reduction up to 6X Capacity Savings

NeuralMesh 6 also enables data reduction, including fingerprinting, similarity hashing, deduplication, and compression, by default across every deployment, with a write overhead below 5%, up to 6x capacity savings on AI training data, and a contractual guarantee covering both reduction ratio and performance impact. A new Kubernetes Operator automates cluster deployment and lifecycle management for organizations running Kubernetes as their standard operating model, which WEKA says cuts deployment time from weeks to hours. NeuralMesh Observe, included at no extra cost with every deployment, provides SaaS-based multi-cluster dashboards, client-level diagnostics, and alerting routed to Slack, PagerDuty, or email.

WEKA reports using its Augmented Memory Grid feature in production, which expands GPU memory by speeding up persistent KV cache access to NeuralMesh-managed NVMe storage on Oracle Cloud Infrastructure. Benchmarks on OCI H100 infrastructure demonstrated 10x higher token throughput, 10x more concurrent users served, and 7x more tokens per GPU, which WEKA says is measured against DRAM-based alternatives. Pablo Selem, senior director of software development at OCI, characterized the approach as removing memory bottlenecks so customers can achieve higher throughput and more users from the same GPU footprint.

WEKApod 3: Custom Hardware Built Around the Software

WEKApod 3 is WEKA’s own hardware design rather than a reference architecture built on third-party OEM chassis. The company says a single WEKApod rack delivers 1.1 exabytes of effective capacity on a hardware foundation of 441.5 PB of raw capacity, making it the first single-rack system to exceed an exabyte of effective capacity. Per-rack throughput is rated at 10.2 TB/s with 210 million IOPS. WEKA reports 267% higher effective capacity density and 114% higher throughput density per rack unit than the next-best publicly available alternative in each category.

The design relies on a PCIe Gen 6 internal fabric, a cable-based drive interconnect rather than a backplane, NVIDIA ConnectX SuperNIC networking for Spectrum-X Ethernet connectivity, and a software-managed thermal architecture rated for 35°C ambient operation that throttles NVMe power under thermal stress instead of shutting down. WEKA has multiple patents pending on the chassis, drive interconnect, thermal management, and serviceability design. Serviceability features include hot-pluggable boot drives with a GUI-guided replacement process WEKA says takes about 10 minutes instead of a multi-hour maintenance window, along with headless, cloud-driven rack-scale deployment through NeuralMesh Home.

Three configurations for different workload priorities

WEKApod Nitro is designed for workloads where storage bandwidth is critical, ensuring GPUs remain saturated. It features a two-rack-unit, four-node chassis with four independent failure domains and 56 TLC drives, supported by dual-port NVIDIA ConnectX networking that delivers 800 Gb/s throughput. WEKApod Prime emphasizes balanced capacity and performance with an AlloyFlash blend of TLC and QLC drives, housed in a similar four-node, two-rack-unit chassis that supports 56 drives. WEKApod Prime Max maximizes capacity in a compact form: a two-rack-unit, two-node chassis containing 70 NVMe drives, using Micron’s 245.76 TB 6600 ION SSDs, combined with NeuralMesh’s object storage and data reduction techniques, to achieve an effective capacity of 1.1 exabytes in a single 56U rack.

AlloyFlash, the tiering feature that makes the Prime and Prime Max configurations viable, automatically routes latency-sensitive operations to TLC flash while directing bulk-capacity data to QLC, which runs roughly 30-40% cheaper per terabyte, without requiring customer configuration. This is the clearest point at which the NeuralMesh 6 software release and the WEKApod 3 hardware release are one product: the software’s tiering logic makes the higher-density hardware configurations usable at production performance levels, rather than just a larger capacity number on a spec sheet.

WEKA bases its hardware decision on current data center constraints: US data center construction dropped in 2025 for the first time since 2020, grid connection queues in major markets now take four to seven years, and Morgan Stanley forecasts a 49-gigawatt power shortfall in the US through 2028. This is in addition to ongoing NAND supply issues and longer OEM lead times. WEKA argues that storage that is inefficient in rack space and power directly competes with GPUs for limited physical resources. By managing its own hardware supply chain rather than relying on OEM channels, it believes it can offer more predictable pricing and lead times for customers planning large-scale infrastructure projects.

Jason Hardy, VP of Storage Technology at NVIDIA, said Spectrum-X Ethernet networking gives WEKApod 3 the high-bandwidth, low-latency fabric needed to keep the storage-to-GPU data path clear at scale. Steve McDowell, chief analyst at NAND Research, argued that inference at production scale is a different infrastructure problem than training, with tokens per rack, tokens per watt, and cost per inference at sustained load becoming the metrics that matter, a scorecard he says buyers should evaluate every vendor against. Jeremy Werner, senior vice president and general manager of Micron’s Core Data Center Business Unit, added that the new WEKApod architecture with Micron’s 245TB SSDs delivers 15.8 petabytes in a 2U footprint, preserving power and space for additional compute.

Availability

NeuralMesh 6 is expected to be generally available in the second half of 2026. Current WEKA customers can upgrade at no additional cost through standard channels. WEKApod Nitro, Prime, and Prime Max are now available for ordering through WEKA’s distributor and VAR network, with deliveries starting in fall 2026 and NeuralMesh 6 pre-installed. Additionally, WEKApod is now offered in configurable SKUs for the first time in this generation, allowing customers to choose chassis type, memory, drive capacity, and drive count, supporting systems from under 1PB up to 100PB or more in a single setup.

The post WEKA’s WEKApod 3 Breaks the Single-Rack Exabyte Barrier as NeuralMesh 6 Goes Multi-Tenant appeared first on StorageReview.com.

Dell Technologies closed the first quarter of 2026 as the top external enterprise storage vendor worldwide, according to IDC’s Worldwide Quarterly Enterprise Storage Systems Tracker released in June, posting 31.2% market share on 40.8% year-over-year growth, nearly double the 22.7% pace of the overall segment. By IDC’s count, the rest of the top five vendors combined did not outsell Dell for the quarter. While the share figure is the headline, Dell’s broader message is that the result reflects continued execution around three customer priorities: private cloud flexibility, AI infrastructure built on enterprise-owned data, and cyber resilience embedded into the platform.

That positioning is especially relevant for Dell because the enterprise storage conversation has shifted beyond raw capacity and performance. Customers increasingly want infrastructure that supports mixed hypervisor environments, AI data pipelines, and faster recovery from cyber events without requiring wholesale architectural changes. Dell’s current portfolio, with PowerStore at the center of the midrange discussion, is being framed around that convergence.

On the private cloud side, Dell is targeting enterprises that want both operational simplicity and infrastructure choice. Through Dell Private Cloud and the Dell Automation Platform, customers can deploy cloud software stacks from Broadcom, Microsoft, Nutanix, and Red Hat on disaggregated Dell infrastructure rather than being tied to a single hyperconverged model. At Dell Technologies World, the company added support for VMware Cloud Foundation 9.1, Microsoft Azure Local, and PowerStore integration with Nutanix AHV.

The practical value of that approach is that compute and storage can scale independently while lifecycle management remains automated. Dell is also making a cost argument, saying this model can reduce costs by up to 65% compared with traditional HCI approaches. For infrastructure teams trying to balance modernization with budget control, that flexibility is a meaningful part of the story.

PowerStore remains a key piece of that strategy. Dell is positioning PowerStore Elite as the storage foundation for organizations that need to consolidate block, file, virtual machine, and container workloads on a single platform while maintaining hypervisor flexibility. The company says the platform delivers up to three times the performance of prior models, includes a 6:1 data reduction guarantee, and supports non-disruptive upgrades. Existing customers can also modernize through mixed-generation clustering, which allows newer and older systems to operate together and reduces the operational friction that often comes with hardware refresh cycles.

That upgrade path is one reason PowerStore continues to resonate in the market. It gives customers a way to evolve their infrastructure without forcing a hard cutover while keeping pace with changing workload demands across virtualized environments, databases, and increasingly containerized applications. For a platform with broad enterprise deployment, continuity matters as much as peak performance.

Dell is making a similarly integrated case around AI. The company’s view is that successful enterprise AI initiatives depend less on public data sources and more on the ability to discover, govern, and operationalize internal data. Its AI Data Platform, part of the broader Dell AI Factory framework, is designed to index billions of unstructured files and orchestrate them through managed data pipelines to move customers from AI pilots into production more quickly.

Dell also highlighted GPU-accelerated analytics, noting that the platform can deliver SQL query performance up to 6 times faster when paired with NVIDIA Blackwell GPUs. Underneath that software layer, storage remains central. PowerScale and ObjectScale are positioned as the systems that turn fragmented enterprise data into AI-ready repositories. Dell said the new ObjectScale X7700 ultra-dense appliance offers up to 45% more HDD capacity than its predecessor, improving economics for large-scale object storage deployments. The company also said support for 245TB all-flash drives is coming, which would more than triple ObjectScale’s flash density.

Cyber resilience is the third pillar of Dell’s current storage narrative and is increasingly tied to both private cloud and AI deployments. Enterprise customers are no longer treating cyber recovery as a secondary planning exercise. Instead, the expectation is that ransomware detection, protection orchestration, and large-scale recovery need to be part of the platform itself.

Dell is addressing that with PowerProtect One, which brings data protection orchestration and storage under a single control plane. The company said this can reduce management overhead by up to 50% while preserving large-scale recovery capabilities and established data reduction efficiency. It also pointed to Cyber Detect, which integrates AI-based ransomware detection with PowerStore and PowerMax; Dell cites a 99.99% detection confidence from a third-party validation of the underlying Index Engines technology, with the goal of initiating recovery operations as soon as suspicious activity is detected.

Taken together, these elements help explain why Dell’s storage message is landing with enterprise buyers. Rather than presenting private cloud, AI infrastructure, and cyber recovery as separate product discussions, Dell is tying them back to a common infrastructure foundation. In that context, PowerStore plays an important role because it sits at the intersection of consolidation, modernization, operational simplicity, and resilience.

For customers already invested in Dell infrastructure, that consistency reduces friction. For prospective buyers, it strengthens the case that Dell’s storage portfolio is not just broad, but aligned with how enterprise IT priorities are evolving. The Q1 share gains and growth numbers may capture attention. Still, the more durable story is that Dell continues to pair scale with a portfolio strategy that maps cleanly to current enterprise requirements.

The post Dell Outsells the Rest of IDC’s Top Five Combined in Q1 External Storage appeared first on StorageReview.com.

Scality and WEKA have expanded their partnership in France with a new joint customer support agreement that gives French organizations a single in-country point of contact for sales and tier-one technical support. The agreement, announced July 7 ahead of RAISE Summit 2026 in Paris, is intended to simplify deployment and ongoing operations for enterprises adopting AI infrastructure.

Under the agreement, Scality will distribute the joint solution to customers in France, provide local front-line support, and promote the offering throughout the country. The companies are targeting enterprise customers, government organizations, and AI developers that require high-performance storage combined with cyber resilience and data sovereignty.

Integrated Solution

The integrated platform combines WEKA NeuralMesh with Scality Autonomous Data Infrastructure (ADI). WEKA positions NeuralMesh as a high-performance AI storage and context memory platform designed to improve storage throughput and maintain high GPU utilization during AI training and inference workloads. Scality ADI complements the platform by providing autonomous data management, cyber resilience, and sovereign data control across the entire data lifecycle, from active datasets to long-term object storage at exabyte scale.

The expanded support agreement builds on the jointly validated architecture the companies announced on February 24, 2026, which pairs WEKA NeuralMesh with a cost-efficient Scality RING object tier. The integration uses Scality’s lightweight object connector for NeuralMesh to move data efficiently between high-performance and capacity storage tiers. According to Scality’s testing, the combination delivers up to 10x faster performance and up to 20% lower infrastructure costs.

Local-Language Support

Beyond technical integration, the companies are also increasing their joint go-to-market activities in France. By combining local sales efforts with French-language technical support, Scality and WEKA aim to reduce deployment complexity and provide a single support path for customers implementing AI infrastructure in regulated environments.

Scality CEO Jérôme Lecat said AI deployments are increasingly constrained by data infrastructure rather than GPU availability. He said the partnership combines the strengths of both platforms while adding local French-language support, providing customers with a unified point of contact for mission-critical deployments.

WEKA Chief Strategy Officer Nilesh Patel said AI adoption is often limited by infrastructure performance rather than model availability. He noted that NeuralMesh is designed to maximize GPU utilization and inference throughput while reducing cost per token. At the same time, Scality provides cyber resilience and sovereign data controls required by many regulated European organizations. He added that local support helps reduce operational risk and accelerate production deployments.

The post Scality and WEKA Expand Partnership in France with Local AI Infrastructure Support appeared first on StorageReview.com.

At the RAISE Summit in Paris, DDN introduced Infinia 2.4, the latest release of its AI data platform designed for production AI, large-scale inference, and sovereign AI deployments. The update expands enterprise capabilities with multi-tenancy, governance, identity management, and new POSIX support while continuing to optimize data access for GPU-accelerated AI infrastructure.

As enterprise AI deployments mature beyond model training, infrastructure priorities are shifting toward improving inference efficiency, reducing cost per token, and maximizing GPU utilization. DDN positions Infinia 2.4 as an enterprise data platform intended to help organizations operate AI factories more efficiently by reducing storage bottlenecks and simplifying operations across shared environments.

The release also enhances NVIDIA DSX-based AI factory deployments, referencing NVIDIA’s Omniverse DSX blueprint for gigascale AI factories, by improving data throughput and GPU utilization while streamlining infrastructure management. DDN says these improvements can accelerate inference performance and improve the return on the billions of dollars organizations are investing in GPU infrastructure.

DDN CEO and Co-Founder Alex Bouzari said the industry focus is moving from GPU acquisition toward operational efficiency metrics such as inference performance, GPU utilization, and cost per token. He added that enterprise AI deployments require infrastructure that combines performance, governance, and security to maximize the value of AI investments.

Focus on Inference Performance

Inference has become the dominant operational cost for many enterprise AI deployments as organizations expand the use of retrieval-augmented generation (RAG), AI copilots, agentic AI, and autonomous applications. Infinia 2.4 is designed to improve inference efficiency through low-latency data access, high-performance object storage, and data services intended to keep accelerators fully utilized.

Key platform enhancements include:

• High-performance distributed KV cache acceleration
• Sub-millisecond access to AI datasets and model artifacts
• High-concurrency support for multi-tenant inference environments
• Optimizations for RAG, vector databases, agentic AI, and large-scale inference workloads
• Improved GPU utilization to reduce infrastructure overhead

DDN says these capabilities are intended to improve response times while lowering the cost per token for production AI services.

Expanded Enterprise Capabilities

A significant focus of Infinia 2.4 is support for shared enterprise AI infrastructure. The platform adds capabilities DDN says are required by many of NVIDIA’s cloud partners, managed AI service operators, and enterprise AI platforms, including advanced multi-tenancy, identity integration, quota enforcement, governance controls, and stronger operational isolation between tenants.

These additions allow multiple organizations, business units, or sovereign AI environments to share a common infrastructure while securely maintaining operational separation.

Initial POSIX Support

Infinia 2.4 also introduces limited availability POSIX support, expanding application compatibility beyond object storage. The initial release includes qualified POSIX clients for Red Hat Enterprise Linux and Ubuntu, documented deployment guidance, and defined throughput commitments. The feature enables additional AI and data-intensive workloads while maintaining the platform’s existing performance architecture.

S3 Compatibility

Existing deployments can upgrade to Infinia 2.4 without modifying applications built around Amazon S3 APIs. DDN says the platform maintains compatibility with current S3 environments and SDKs while providing the latest performance and scalability improvements.

Broader AI Infrastructure Strategy

The release aligns with DDN’s continued focus on enterprise AI, hyperscale infrastructure, inference platforms, and sovereign AI initiatives. According to the company, its technology is deployed across AI infrastructure supporting organizations including NVIDIA, xAI, Salesforce, Mistral, SK Telecom, Yotta, and multiple government and research institutions, where it is used to improve resource utilization, accelerate model deployment, and support large-scale AI operations.

The post DDN Infinia 2.4 Adds Multi-Tenancy and POSIX Support for Production AI Factories appeared first on StorageReview.com.

Canonical has released LXD 6.9 with updates aimed squarely at storage teams: a native driver for Dell PowerStore arrays, a new Fibre Channel connector for remote storage generally, and support for Dell PowerFlex 5. For a platform that started life as a container manager, that is a notable amount of enterprise SAN plumbing in a single release.

For readers who have not tracked it, LXD is Canonical’s open-source virtualization platform that manages both system containers and full KVM virtual machines across clustered hosts. It has gained attention as organizations reassess their hypervisor options in the wake of Broadcom’s VMware licensing changes, and Canonical has been steadily building out the enterprise features (clustering, live migration, a Kubernetes CSI driver, disaster-recovery replication) that a VMware alternative needs. The missing piece for many has been the storage they already have, and this release addresses much of it.

The reason a native driver matters comes down to where instance data lives. Without one, LXD typically puts VM and container volumes on host-local ZFS, LVM, or Btrfs, or on a Ceph cluster, which means an existing array is reduced to serving LUNs that the host then carves up itself. With a native driver, LXD provisions each instance volume directly on the array, so snapshots, clones, and thin provisioning are handled by the array’s own data services, and volumes are reachable from any cluster member. PowerStore now joins Dell PowerFlex, Pure Storage, and HPE Alletra on that list, with both iSCSI and Fibre Channel connectivity supported at launch.

The Fibre Channel connector is arguably the bigger long-term change. Until now, LXD’s remote storage drivers supported only iSCSI or NVMe/TCP, which excluded the large installed base of FC fabrics that dominate legacy SAN estates. The connector is a general transport layer, so drivers beyond PowerStore can adopt it. In related housekeeping, the NVMe/TCP pool mode has been renamed from nvme to nvme/tcp, with existing pools migrated automatically on upgrade.

On the PowerFlex side, the driver now supports PowerFlex 5, including thin clone support, and automatically detects the array’s software version while remaining compatible with PowerFlex 4. The ZFS driver also gains a practical speedup: LXD now caches image variants matching an instance’s configuration, so repeated deployments from the same image no longer rebuild the clone.

Beyond storage, 6.9 adds load balancer pools for OVN networks with health checking, OWASP-compliant security event logging that can be routed to Grafana Loki, and quorum protection for cluster evacuations. The release also lands fixes for eleven CVEs, several of which involved symlink attacks in crafted images or project restriction bypasses.

One caveat: 6.9 is a feature release, which Canonical explicitly does not recommend for production use. Shops that want these capabilities on a supported footing will be waiting for the next LTS. For everyone else, the release is available now via snap install lxd --channel=6/stable, with the snap base moving from core24 to core26.

The full release notes are available on Canonical’s LXD documentation site.

The post Canonical LXD 6.9 Adds Dell PowerStore Driver and Fibre Channel Support appeared first on StorageReview.com.

NetApp has introduced StorageGRID 12.1, adding a federated global namespace and a range of performance, data management, and security enhancements aimed at AI, analytics, and other large-scale object storage workloads. The release is designed to simplify the management of globally distributed object data while improving the efficiency of AI data pipelines and modern cloud-native applications.

As enterprises continue to generate rapidly expanding volumes of unstructured data, object storage platforms are increasingly expected to serve as the foundation for AI infrastructure in addition to traditional backup and archival use cases. StorageGRID 12.1 addresses this shift by improving access to distributed data and streamlining operations across hybrid environments.

A key addition in StorageGRID 12.1 is the Global Federated Namespace, which allows organizations to manage multiple geographically distributed StorageGRID deployments through a single namespace. NetApp says the capability scales to up to 10 exabytes without requiring applications or existing workflows to be redesigned, making it easier to build globally distributed AI data lakes and object repositories.

The release also delivers significant performance improvements over the previous version. According to NetApp, StorageGRID 12.1 can provide up to 400% higher throughput than StorageGRID 12.0, depending on workload characteristics and object size. At scale, the platform can deliver up to 12 TB/s of aggregate throughput for large AI infrastructure deployments.

Operational improvements focus on managing increasingly large object repositories. New batch operations enable administrators to perform actions across billions of objects, while enhanced change tracking allows AI applications and agents to identify modifications made to object storage buckets since a previous scan. These capabilities are intended to reduce the overhead associated with maintaining AI data pipelines and keeping large datasets synchronized.

Security and governance have also been expanded in StorageGRID 12.1. Multi-admin verification introduces additional administrative controls for organizations operating in regulated industries, helping enforce governance policies around critical configuration changes.

NetApp said the new release extends its data platform by providing a unified global namespace that enables organizations to manage large-scale distributed datasets and accelerate AI and analytics workloads, regardless of where data resides.

Separately, NetApp announced it has been named a Leader in The Forrester Wave: Object Storage Solutions, Q2 2026. In its inaugural evaluation of the object storage market, Forrester cited NetApp’s strategy for hybrid, multicloud, and sovereign deployments, noting the platform’s suitability for enterprises managing distributed and regulated object storage environments while supporting AI-focused storage services.

The post NetApp StorageGRID 12.1 Adds a Federated Namespace and Up to 12 TB/s for AI Workloads appeared first on StorageReview.com.

At ISC High Performance 2026 in Hamburg, VDURA is showcasing three major platform updates designed to improve storage operations and data pipeline performance for AI and HPC environments. The company unveiled a next-generation multi-tenant control plane, enhanced S3 performance capabilities, and native S3 object tagging support, all scheduled for general availability in the second half of 2026.

The updates target organizations that manage large-scale AI training, inference, and scientific computing workloads, where operational simplicity and sustained storage performance are critical.

New Control Plane Targets Multi-Tenant AI and HPC Deployments

VDURA’s next-generation control plane introduces a redesigned management experience focused on multi-tenant environments. The platform adds a modernized management interface and a simplified tenant administration model, allowing platform administrators and tenant operators to manage storage resources from a centralized dashboard.

The release also includes a REST API for key platform operations, providing integration points for automation frameworks, orchestration tools, and operational workflows. Together, the interface and API are intended to reduce management complexity while maintaining the level of control required by enterprise, cloud, and research deployments.

S3 Performance Optimized for AI Data Pipelines

VDURA is also delivering a series of S3 performance enhancements aimed at cloud-native AI workflows. The company said the updates are designed to sustain high throughput across data-intensive operations such as model checkpointing, inference serving, and large-scale dataset ingestion.

The improvements focus on reducing latency for S3-native operations while increasing aggregate throughput across concurrent read and write workloads. As AI environments continue to scale in both dataset size and user concurrency, maintaining consistent object storage performance has become increasingly important for minimizing pipeline bottlenecks.

Native S3 Object Tagging Adds Metadata-Based Data Management

Complementing the performance updates, VDURA is introducing native S3 object tagging support. The capability allows organizations to associate metadata with stored objects, creating a framework for policy-driven data management and governance.

With object tagging, administrators can implement lifecycle policies, automate tiering workflows, and apply more granular access controls across datasets, model artifacts, and scientific research repositories. For organizations operating petabyte-scale object storage environments, metadata-driven management can help streamline governance and data lifecycle processes without requiring changes to application workflows.

Customer-Driven Enhancements

“With these new capabilities, we’re focused on making VDURA more powerful and easier to operate at every level of the organization,” said Chris Girard, Vice President of Product Management at VDURA. “The new management interface brings clarity and control to the people running VDURA day-to-day, while our expanded S3 capabilities give data engineers and AI practitioners the tools to build more sophisticated, automated data pipelines. These are the operational and integration capabilities our customers have been asking for.”

Availability

The next-generation control plane, S3 performance enhancements, and native S3 tagging capabilities are expected to become generally available during the second half of 2026 for all V5000-class systems. Existing customers will be able to deploy the new capabilities through an in-place online software update.

The post VDURA Unveils Multi-Tenant Control Plane and S3 Enhancements at ISC 2026 appeared first on StorageReview.com.

CGI and NetApp have expanded their global alliance partnership, with NetApp Keystone set to power CGI’s block storage solutions within its shared services platform. The move deepens the companies’ existing relationship and is designed to help organizations modernize IT infrastructure, improve data management, and advance AI initiatives across private, public, and hybrid cloud environments.

Under the expanded agreement, CGI will integrate NetApp Keystone’s subscription-based storage model into its shared services platform. The offering is designed to provide customers with scalable block storage resources while allowing organizations to consume storage capacity through a flexible, consumption-based model rather than traditional infrastructure procurement cycles.

NetApp Keystone delivers storage services across on-premises and cloud environments, combining performance, data management capabilities, and high-availability features. The platform also incorporates integrated security features focused on threat detection, data protection, and recovery, supporting organizations with increasingly stringent operational and cybersecurity requirements.

The partnership combines NetApp’s storage and data management portfolio with CGI’s consulting, cloud, AI, and managed services expertise. NetApp is one of more than 150 technology firms in CGI’s Global Alliances network, and together the companies are targeting enterprises looking to modernize infrastructure while maintaining operational consistency across hybrid environments.

“The expansion of our partnership with NetApp reflects a strong commitment on both sides to drive meaningful outcomes for our clients,” said Virginia Williams, Senior Vice-President and Business Unit Leader, U.S. Northwest Operations at CGI. “The technology, expertise and innovation offered by this powerful alliance will continue to help clients modernize their IT environments, become more data-driven and prepare for AI at scale.”

This next phase of the alliance will see the two companies working together to design, deliver, and operate solutions to meet clients’ evolving digital needs. CGI will deliver services on behalf of NetApp, while NetApp will partner with CGI to deliver enterprise-grade data and storage services that enable flexible, consumption-based solutions for joint clients across industries.

“By expanding our partnership with CGI, we’re enabling our shared customers to build a resilient, secure solution that delivers consistent performance and intelligent data management for their most critical workloads,” said Alvaro Celis, Chief Partner and Ecosystem Officer at NetApp. “Working side-by-side, CGI and NetApp will continue to empower organizations to achieve better business outcomes through an intelligent data infrastructure that simplifies hybrid cloud adoption and securely unlocks greater value from their data.”

The announcement reflects a broader enterprise trend toward storage-as-a-service consumption models, particularly as organizations seek greater operational flexibility while preparing infrastructure for data-intensive AI workloads and hybrid cloud deployments. By incorporating Keystone into its shared services platform, CGI adds a scalable storage foundation that can be delivered as part of larger managed services and digital transformation engagements.

The post CGI Taps NetApp Keystone to Power Block Storage in Its Shared Services Platform appeared first on StorageReview.com.