# Storage News Ticker - 2 July 2026 — retained record

> Retained news record from news:blocksandfiles.

Canonical: https://highsignal.app/data/records/02c43efad353ea6b

[← sources](https://highsignal.app/data)


source: [news](https://highsignal.app/data/news) (news:blocksandfiles)  published 2026-07-02 17:19 ingested 2026-07-03 09:03

[original source ↗](https://www.blocksandfiles.com/flash/2026/07/02/storage-news-ticker-2-july-2026/5265934)

## extract

Storage News Ticker - 2 July 2026
Japan’s NTT DOCOMO is using Dell PowerMax and PowerStore, paired with APEX Subscriptions and Dell Managed Services behind in its internal private cloud supporting its SmartLife services wit 100 million users. It gained:
50%+ reduction in seven-year infrastructure costs
40% less time spent on infrastructure operations
PowerMax hit 6:1 to 8:1 data reduction in production, cutting rack space and power nearly 50%
Cost per GB down roughly half versus the prior environment
Maintenance coordination time down more than three hours per instance
Dell is migrating DOCOMO to its PowerMax and Dell PowerStore-based core infrastructure in phases, with full completion expected by the end of 2026. Based on a seven-year cost comparison, the combined platform, APEX consumption model and Dell Managed Services are projected to reduce long-term infrastructure costs by more than 50 percent.
…
Nvidia and IBM say analytical SQL is a great fit for GPUs: scans and aggregations over columnar Parquet data are massively parallel and bandwidth-hungry, precisely the workload modern data-center GPUs are built for. They have brought GPU-accelerated operators to the Velox execution engine that powers Presto C++, and built a new UCX-based exchange that exploits fast interconnects like NVLink. This setup can run analytical SQL much faster than CPUs: published numbers show a single node Presto C++ with GPU-accelerated operators running a TPC-H-style benchmark in ~100 seconds versus ~1,200 seconds on a high-end CPU — on the order of 12× faster — and a UCX/NVLink exchange running >6× faster with multi-node Presto.
…
Denmark-based Keepit which provides a backup storage facility in its own cloud, announced the renewal and upsizing of its credit facilities to $90 million, with continued support from the Export and Investment Fund of Denmark (EIFO) and HSBC Innovation Banking. The refinancing maintains total facilities at $90 million, consolidating and extending terms to align with Keepit's growth roadmap. Building on the momentum of its $50 million equity raise in December 2024 and the prior $60 million upsizing announced in September 2025, the proceeds will support product innovation ― in particular the buildout of Keepit's emerging AI service portfolio ― as well as expanded go-to-market capacity and deepened penetration of the enterprise segment.
…
At the IEEE/JSAP VLSI Symposium 2026, Kioxia and Sandisk presented the world’s first successful QuadLevel Cell (QLC) operation of a multi-stacked cell array CMOS directly bonded to array (MSA-CBA). This breakthrough overcomes key challenges in high stacking of 3D flash memory: cell current degradation, wafer warpage, and large block (BLK) size. These results mark a milestone toward ultra-high-density 3D flash memory with over 1,000 stacked layers.
…
The memory shortage is expected to get worse, with TF International analysts expecting 15-20% of capacity allocated to consumer electronics shifting to data centers in 2027. Gui Soubihe, CEO of Latitude.sh thinks that consumers would inevitably feel the knock-on effects in the form of not only higher prices, but canceled products, especially the budget lineups. Laptop makers, for instance, might have to stop making their cheapest models as individual component costs skyrocket.
…
The EX-5410C is Liqid's implementation of a memory pooling fabric across CXL and LIqid says it can solve a GPU server’s memory wall by pooling the server’s CPU DRAM. It orovides:
Up to 100 TB of DRAM pooled in a single deployment
Shared across as many as 32 server nodes
Roughly 200 nanoseconds access latency, near-native DRAM speed
Near native-DRAM speed and access latency
When a GPU node needs 6 TB for a long-context inference job, it gets 6 TB from the pool, and when the job completes, the capacity returns for the next allocation. Liqid Matrix handles the orchestration, with Kubernetes, Slurm, OpenShift, and Ansible integrations. Liqid Memory Pooling is shipping in production today. The CXL consortium's 3.0/3.1 roadmap targets multi-rack memory pooling for late 2026-2027, which extends the same architecture from rack scale to row scale.
…
OceanBase released OceanBase AI Database, designed to enable enterprises to manage multimodal data, deliver real-time trusted data context to AI agents, and simplify fragmented data architectures. OceanBase is an enterprise-grade, open-source distributed SQL database initially developed in 2010 by Ant Group and Alibaba to handle massive transactional (OLTP) and analytical (OLAP) workloads. It features include:
OceanBase Lakebase: Serving as the underlying data engine, Lakebase enables structured, unstructured, and vector data to be managed, processed, searched, and served within a unified architecture.
OceanBase DataStudio: Built upon the Lakebase engine, DataStudio is a data production, governance, and services platform, covering the entire lifecycle from data ingestion, processing, and orchestration to semantic modeling and agent collaboration, and turning data silos into callable data services.
OceanBase DataPilot: Acting as a business intelligence agent for enterprises, DataPilot can generate analytical reports, data dashboards, and trusted answers through natural language, making data intelligence highly accessible to non-technical users.
Check out this Forrester multi-model database report for more information about OceanBase.
…
Nasuni earned its sixth consecutive NorthFace ScoreBoard Service Award (NFSB) from the Customer Relationship Management Institute LLC (CRMI) and securing top rankings across 15 categories in G2’s Summer 2026 Reports.
…
Pinecone Nexus, a knowledge engine designed to connect AI applications to reliable, real-time data while powering operational and infrastructure costs, is officially in public preview. An enterprise’s business context lives distributed across contracts, wikis, HR docs, meeting notes, support tickets, and financial records. Nexus compiles this distributed knowledge into a structured layer agents can query directly, shifting token spend out of the per-query retrieval loop and into a one-time curation step. Nexus is built for bounded corpora where agents need to reason across documents. It's at its best on the kind of corpus where a single question touches dozens of files and standard retrieval starts falling apart. Read more in a blog.
This is context memory and several database suppliers, such as Couchbase, HYCU and Redis, are recognizing a need for it and storing it for AI agent access.
…
Rubrik ZeroLabs research found four Langflow vulnerabilities;
Unauthenticated RCE via public chatbot route (CVE-2026-48519, Critical 9.6): When a user interacts with a publicly shared Langflow chatbot, the platform sends the full graph definition — including every component's Python code — to the server for execution. This effectively creates a "send me Python code and I'll run it" endpoint, accessible to any unauthenticated user.
"RAGPull" — file read to full server takeover (CVE-2026-7524, Critical 9.8): Langflow's document parser, used to feed files into RAG-based chatbots, supports tar archive uploads. Because tar files can contain symlinks, an attacker can upload a malicious archive pointing to sensitive server files — including the environment variables that store Langflow superuser credentials — extract those credentials via the chatbot, and use them to execute arbitrary code on the server.
Unauthenticated file upload / storage DoS + path disclosure (CVE-2026-7528, High 7.1): An unauthenticated upload route allows any attacker with network access to fill Langflow's storage and cause a denial of service. The route also returns the full absolute file path of uploaded files, leaking server information.
Local file read via RAG parser (CVE-2026-48520, Moderate 6.1): A separate field in the public chatbot route allows local files from the Langflow filesystem to be fed directly into the LLM, with potential for further exploitation.
They were communicated to the Langflow team and have been patched.
…
Samsung and Google are involved in NVMe SSD virtualization via NVM Express TP4193: PCIe Exported NVM Subsystem Migration. This introduces new standardized virtual objects (exported) and redefines how virtual storage is created, managed, and moved.Conventionally, SSD storage areas are allocated to virtual machines using a direct assignment mechanism, typically implemented in the form of logical partitioning at the PCIe level. This can be realized either by the SSD exposing several PCIe physical functions (PFs) or, more recently, through the Single-Root I/O Virtualization (SR-IOV) Extended Capability in which the SSD is logically partitioned into a single physical function and several virtual functions (VFs). To a host, the NVMe controller interface presented by a VF is indistinguishable from one presented by a PF. Direct assignment allows the virtual machine manager (VMM) to map the PCI configuration space of the VF into the address space of the VM, which allows the VM to access the PCIe function as if it were locally attached.
Because each NVMe register is passed into the VM as-is, the VM will also observe the SSD as-is, exposing its identity, unique identifiers and NVM subsystem controllers and namespaces. This is not optimal. With TP4193 the SSD itself exposes logical, virtualized storage constructs with definable behavior. The host becomes an “orchestrator” rather than an implementer. TP4193 has 2 distinct functionalities;
Standardized virtual (exported) object creation: Establishes a common framework for creating and managing exported objects; and
Reporting and capability masking: Defines how a virtual SSD presents, reports, and masks its capabilities and attributes to appear as an independent physical device to the host.
Learn more in a Samsung blog.
…
Object storage supplier Scality and French public cloud provider OVH are partnering to provide a joint cloud storage platform providing European digital sovereignty. Scality RING and ARTESCA provide massive, resilient, S3-compatible object storage deployed directly on the customer’s on-premises infrastructure. OPCP (On-Prem Cloud Platform), adds cloud-native orchestration, a managed services catalogue, and deployment automation, all built on a foundation independent of foreign public cloud providers. Deployment options include a 100% dedicated sovereign cloud and on-premises via OPCP with backup replication across multiple availability zones (multi-AZ). IT also includes a dedicated storage offering on OVHcloud Bare Metal infrastructure through the HGR-STOR range.
Customers can deploy off-site backups with private S3 compatibility, integrated with leading cyber-resilience tools such as Veeam and Commvault, all hosted directly by OVHcloud. There is a guarantee of 100% dedicated servers, with no resource sharing with other customers, providing predictable performance and maximum isolation. It also supports a robust business continuity plan (BCP) by enabling controlled failover to OVHcloud infrastructure in the event of a disaster or primary site outage.
The joint Scality-OVHcloud OPCP platform supports AI workloads. Organisations can train models, run MLOps pipelines, and deploy AI applications in near real time directly within their own environments, without compromising data security or regulatory compliance. Find out more from OVH.
…
A July 16 SNIA webinar will look at implementing vector search with Ceph object storage. It will explore how LanceDB libraries power a vector search implementation in Ceph, delivered through S3 Vectors API actions. Attendees will see how Ceph can provide out of the box, billion scale, multi tenant nearest neighbor search that meets the needs of most RAG workloads while preserving the unified, software defined storage model enterprises already trust. Register for the event here.
…
Walrus will be available to the 2,000+ startups and protocols building on Lighthouse, an IPFS (InterPlanetary File System) developer platform. Through the integration, developers can access Walrus' verifiable storage and performance benefits without changing their existing CID-based (Content Identifier-based) workflows, SDKs, or gateways.

## retained text (truncated)



```
Storage News Ticker - 2 July 2026
Japan’s NTT DOCOMO is using Dell PowerMax and PowerStore, paired with APEX Subscriptions and Dell Managed Services behind in its internal private cloud supporting its SmartLife services wit 100 million users. It gained:
50%+ reduction in seven-year infrastructure costs
40% less time spent on infrastructure operations
PowerMax hit 6:1 to 8:1 data reduction in production, cutting rack space and power nearly 50%
Cost per GB down roughly half versus the prior environment
Maintenance coordination time down more than three hours per instance
Dell is migrating DOCOMO to its PowerMax and Dell PowerStore-based core infrastructure in phases, with full completion expected by the end of 2026. Based on a seven-year cost comparison, the combined platform, APEX consumption model and Dell Managed Services are projected to reduce long-term infrastructure costs by more than 50 percent.
…
Nvidia and IBM say analytical SQL is a great fit for GPUs: scans and aggregations over columnar Parquet data are massively parallel and bandwidth-hungry, precisely the workload modern data-center GPUs are built for. They have brought GPU-accelerated operators to the Velox execution engine that powers Presto C++, and built a new UCX-based exchange that exploits fast interconnects like NVLink. This setup can run analytical SQL much faster than CPUs: published numbers show a single node Presto C++ with GPU-accelerated operators running a TPC-H-style benchmark in ~100 seconds versus ~1,200 seconds on a high-end CPU — on the order of 12× faster — and a UCX/NVLink exchange running >6× faster with multi-node Presto.
…
Denmark-based Keepit which provides a backup storage facility in its own cloud, announced the renewal and upsizing of its credit facilities to $90 million, with continued support from the Export and Investment Fund of Denmark (EIFO) and HSBC Innovation Banking. The refinancing maintains total facilities at $90 million, consolidating and extending terms to align with Keepit's growth roadmap. Building on the momentum of its $50 million equity raise in December 2024 and the prior $60 million upsizing announced in September 2025, the proceeds will support product innovation ― in particular the buildout of Keepit's emerging AI service portfolio ― as well as expanded go-to-market capacity and deepened penetration of the enterprise segment.
…
At the IEEE/JSAP VLSI Symposium 2026, Kioxia and Sandisk presented the world’s first successful QuadLevel Cell (QLC) operation of a multi-stacked cell array CMOS directly bonded to array (MSA-CBA). This breakthrough overcomes key challenges in high stacking of 3D flash memory: cell current degradation, wafer warpage, and large block (BLK) size. These results mark a milestone toward ultra-high-density 3D flash memory with over 1,000 stacked layers.
…
The memory shortage is expected to get worse, with TF International analysts expecting 15-20% of capacity allocated to consumer electronics shifting to data centers in 2027. Gui Soubihe, CEO of Latitude.sh thinks that consumers would inevitably feel the knock-on effects in the form of not only higher prices, but canceled products, especially the budget lineups. Laptop makers, for instance, might have to stop making their cheapest models as individual component costs skyrocket.
…
The EX-5410C is Liqid's implementation of a memory pooling fabric across CXL and LIqid says it can solve a GPU server’s memory wall by pooling the server’s CPU DRAM. It orovides:
Up to 100 TB of DRAM pooled in a single deployment
Shared across as many as 32 server nodes
Roughly 200 nanoseconds access latency, near-native DRAM speed
Near native-DRAM speed and access latency
When a GPU node needs 6 TB for a long-context inference job, it gets 6 TB from the pool, and when the job completes, the capacity returns for the next allocation. Liqid Matrix handles the orchestration, with Kubernetes, Slurm, OpenShift, and Ansible integrations. Liqid Memory Pooling is shipping in production today. The CXL consortium's 3.0/3.1 roadmap targets multi-rack memory pooling for late 2026-2027, which extends the same architecture from rack scale to row scale.
…
OceanBase released OceanBase AI Database, designed to enable enterprises to manage multimodal data, deliver real-time trusted data context to AI agents, and simplify fragmented data architectures. OceanBase is an enterprise-grade, open-source distributed SQL database initially developed in 2010 by Ant Group and Alibaba to handle massive transactional (OLTP) and analytical (OLAP) workloads. It features include:
OceanBase Lakebase: Serving as the underlying data engine, Lakebase enables structured, unstructured, and vector data to be managed, processed, searched, and served within a unified architecture.
OceanBase DataStudio: Built upon the Lakebase engine, DataStudio is a data production, governance, and services platform, covering the entire lifecycle from data ingestion, processing, and orchestration to semantic modeling and agent collaboration, and turning data silos into callable data services.
OceanBase DataPilot: Acting as a business intelligence agent for enterprises, DataPilot can generate analytical reports, data dashboards, and trusted answers through natural language, making data intelligence highly accessible to non-technical users.
Check out this Forrester multi-model database report for more information about OceanBase.
…
Nasuni earned its sixth consecutive NorthFace ScoreBoard Service Award (NFSB) from the Customer Relationship Management Institute LLC (CRMI) and securing top rankings across 15 categories in G2’s Summer 2026 Reports.
…
Pinecone Nexus, a knowledge engine designed to connect AI applications to reliable, real-time data while powering operational and infrastructure costs, is officially in public preview. An enterprise’s business context lives distributed across contracts, wikis, HR docs, meeting notes, support tickets, and financial records. Nexus compiles this distributed knowledge into a structured layer agents can query directly, shifting token spend out of the per-query retrieval loop and into a one-time curation step. Nexus is built for bounded corpora where agents need to reason across documents. It's at its best on the kind of corpus where a single question touches dozens of files and standard retrieval starts falling apart. Read more in a blog.
This is context memory and several database suppliers, such as Couchbase, HYCU and Redis, are recognizing a need for it and storing it for AI agent access.
…
Rubrik ZeroLabs research found four Langflow vulnerabilities;
Unauthenticated RCE via public chatbot route (CVE-2026-48519, Critical 9.6): When a user interacts with a publicly shared Langflow chatbot, the platform sends the full graph definition — including every component's Python code — to the server for execution. This effectively creates a "send me Python code and I'll run it" endpoint, accessible to any unauthenticated user.
"RAGPull" — file read to full server takeover (CVE-2026-7524, Critical 9.8): Langflow's document parser, used to feed files into RAG-based chatbots, supports tar archive uploads. Because tar files can contain symlinks, an attacker can upload a malicious archive pointing to sensitive server files — including the environment variables that store Langflow superuser credentials — extract those credentials via the chatbot, and use them to execute arbitrary code on the server.
Unauthenticated file upload / storage DoS + path disclosure (CVE-2026-7528, High 7.1): An unauthenticated upload route allows any attacker with network access to fill Langflow's storage and cause a denial of service. The route also returns the full absolute file path of uploaded files, leaking server information.
Local file read via RAG parser (CVE-2026-48520, Moderate 6.1): A separate field in the public chatbot route allows local files from the Langflow filesystem to be fed directly into the LLM, with potential for further exploitation.
They were communicated to the Langflow team and have been patched.
…
Samsung and Google are involved in NVMe SSD virtualization via NVM Express TP4193: PCIe Exported NVM Subsystem Migration. This introduces new standardized virtual objects (exported) and redefines how virtual storage is created, managed, and moved.Conventionally, SSD storage areas are allocated to virtual machines using a direct assignment mechanism, typically implemented in the form of logical partitioning at the PCIe level. This can be realized either by the SSD exposing several PCIe physical functions (PFs) or, more recently, through the Single-Root I/O Virtualization (SR-IOV) Extended Capability in which the SSD is logically partitioned into a single physical function and several virtual functions (VFs). To a host, the NVMe controller interface presented by a VF is indistinguishable from one presented by a PF. Direct assignment allows the virtual machine manager (VMM) to map the PCI configuration space of the VF into the address space of the VM, which allows the VM to access the PCIe function as if it were locally attached.
Because each NVMe register is passed into the VM as-is, the VM will also observe the SSD as-is, exposing its identity, unique identifiers and NVM subsystem controllers and namespaces. This is not optimal. With TP4193 the SSD itself exposes logical, virtualized storage constructs with definable behavior. The host becomes an “orchestrator” rather than an implementer. TP4193 has 2 distinct functionalities;
Standardized virtual (exported) object creation: Establishes a common framework for creating and managing exported objects; and
Reporting and capability masking: Defines how a virtual SSD presents, reports, and masks its capabilities and attributes to appear as an independent physical device to the host.
Learn more in a Samsung blog.
…
Object storage supplier Scality and French public cloud provider OVH are partnering to provide a joint cloud storage platform providing European digital sovereignty. Scality RING and ARTESCA provide massive, resilient, S3-compatible object storage deployed directly on the customer’s on-premises infrastructure. OPCP (On-Prem Cloud Platform), adds cloud-native orchestration, a managed services catalogue, and deployment automation, all built on a foundation independent of foreign public cloud providers. Deployment options include a 100% dedicated sovereign cloud and on-premises via OPCP with backup replication across multiple availability zones (multi-AZ). IT also includes a dedicated storage offering on OVHcloud Bare Metal infrastructure through the HGR-STOR range.
Customers can deploy off-site backups with private S3 compatibility, integrated with leading cyber-resilience tools such as Veeam and Commvault, all hosted directly by OVHcloud. There is a guarantee of 100% dedicated servers, with no resource sharing with other customers, providing predictable performance and maximum isolation. It also supports a robust business continuity plan (BCP) by enabling controlled failover to OVHcloud infrastructure in the event of a disaster or primary site outage.
The joint Scality-OVHcloud OPCP platform supports AI workloads. Organisations can train models, run MLOps pipelines, and deploy AI applications in near real time directly within their own environments, without compromising data security or regulatory compliance. Find out more from OVH.
…
A July 16 SNIA webinar will look at implementing vector search with Ceph object storage. It will explore how LanceDB libraries power a vector search implementation in Ceph, delivered through S3 Vectors API actions. Attendees will see how Ceph can provide out of the box, billion scale, multi tenant nearest neighbor search that meets the needs of most RAG workloads while preserving the unified, software defined storage model enterprises already trust. Register for the event here.
…
Walrus will be available to the 2,000+ startups and protocols building on Lighthouse, an IP
```



document fetched 2026-07-03 09:03 · canonical https://blocksandfiles.com/flash/2026/07/02/storage-news-ticker-2-july-2026/5265934