{"slug": "storage-news-ticker-19-november", "title": "Storage News Ticker - 19 November", "summary": "Airbyte announced semantic search in its context layer, letting AI agents retrieve information such as customer conversations, documents and support tickets and combine it with structured data in Snowflake and BigQuery, available through Airbyte's Model Context Protocol (MCP) server so users can query business data in natural language from Claude, ChatGPT and other MCP-compatible assistants. In the same ticker, Argonaut Labs published an open-source CUDA fork of Salvatore Sanfilippo's ds4 that runs Kimi K3.1 Pro on a single 96GB NVIDIA RTX PRO 6000 Blackwell by streaming roughly 1.45 TB of quantized expert weights from NVMe storage, measuring about 2 tokens/s steady generation (1.94 tokens/s over 512 generated tokens) and about 41 tokens/s prompt processing on a 1,533-token prompt, with 2,048-token context support and reference-model quality validation still outstanding. ClickHouse also acquired Xata, the agent-native Postgres company, adding isolated production-like data environments for developers and AI agents alongside ClickHouse Managed Postgres, Observability, Security and Langfuse's AI Gateway.", "body_md": "# Storage News Ticker - 19 November\n\nConnector supplier **Airbyte** announced semantic search in its context layer, enabling AI agents to find relevant information – such as customer conversations, documents and support tickets – with higher accuracy, and combine those findings with structured data in Snowflake and BigQuery. Available through the [Airbyte Model Context Protocol](https://modelcontextprotocol.io/docs/2026-07-28/getting-started/intro) (MCP) server, the capabilities let users ask business questions in natural language from Claude, ChatGPT and other MCP-compatible AI assistants.\n\n…\n\n**Argonaut Labs** has published an open-source CUDA fork of Salvatore Sanfilippo’s ds4 that runs Kimi K3.1 Pro on one 96GB NVIDIA RTX PRO 6000 Blackwell by streaming approximately 1.45 TB of quantized expert weights from NVMe storage. The published setup uses a three-NVMe RAID 0 array plus two additional NVMe devices holding expert replicas, with host-RAM and GPU caches. Expert placement and read scheduling are central to keeping the GPU supplied. Repeated tests measured approximately 2 tokens/s steady generation, including 1.94 tokens/s over 512 generated tokens. Prompt processing reached about 41 tokens/s on a 1,533-token prompt.\n\nThe implementation is experimental: it currently supports a 2,048-token context, and reference-model quality validation remains outstanding. Get an architecture and benchmark write-up [here](https://argonautlabs.ai/research/kimi-k31-single-gpu.html).\n\n…\n\nCeramic-on-glass archive developer **Cerabyte** announced successful S3 integration testing of its ceramic data storage technology with CERN openlab, demonstrating that the technology can work within existing enterprise and large-scale object storage environments without changes to established workflows. View a summary video - - [here](https://vimeo.com/1242158099/0a7e95685a?share=copy&fl=sv&fe=ci).\n\n…\n\n**Cerabyte’s** Steffen Hellmold will join Gary Grider, Los Alamos National Laboratory, andCandace Culhane, Los Alamos National Laboratory, for a panel discussion at Supercomputing 2026 (SC26) on November 20. They will examine how industry consolidation, limited investment in long-cycle research/development and growing AI and HPC demands are creating new challenges for the future of data storage.\n\n…\n\n**ClickHouse** has acquired Xata, the agent-native Postgres company, giving developers and AI agents isolated, production-like data environments on demand. These updates integrate a transactional database, observability, cybersecurity, and AI engineering into a single, composable platform to simplify operations for enterprises grappling with spiralling AI costs and complexity.\n\nClickHouse is expanding across the four workloads now at the centre of enterprise AI infrastructure. ClickHouse Managed Postgres brings transactional data onto the platform; ClickHouse Observability and ClickHouse Security, built on technology from its September acquisition of RunReveal, bring logs, traces, and security telemetry alongside it; and Langfuse, the open-source AI engineering platform ClickHouse acquired in January, adds an AI Gateway that gives organisations visibility and control over how AI models are used across their applications. All of it runs on the same open, composable foundation as ClickHouse’s analytics engine, so developers and the agents they deploy can build, observe, and secure data-intensive applications in one place, with the cost efficiency to run them at scale.\n\n…\n\nIT and security data engine startup **Cribl** has acquired Tero, which builds an intelligence layer for observability data. Tero co-founders Ben Johnson (CEO), Adam Cohen (CTO), and Nihar Singhal (COO) join Cribl with their team of engineers to help lead its observability roadmap. Johnson previously co-founded Timber, which created the open source Vector observability pipeline and was acquired by Datadog in 2021.\n\n…\n\nCeph-specialist startup **Croit**, based in Munich, is rewriting Ceph in Rust. Ceph is reliable and scales almost without limit, but it was designed when hard disks were the bottleneck. On today's NVMe flash, the software limits performance, not the drive. Rewriting it gives us the chance to build a new architecture for modern hardware, one that gets more out of the hardware people already own and keeps memory use low and predictable. Rust adds memory safety without runtime cost, which matters for software that stores data. The code and the deployment guide are at [https://www.google.com/url?q=http://github.com/croit/ceph-rust&source=gmail&ust=1794403009642000&sa=E](https://www.google.com/url?q=http://github.com/croit/ceph-rust&source=gmail&ust=1794403009642000&sa=E). croit FlashStore OSD is licensed under LGPL-3.0, the Rust client libraries under MPL-2.0.\n\n…\n\n**Databricks** has announced plans to invest approximately $3 billion across Europe and the Middle East over the next three years, its largest regional investment outside North America. The investment will expand Databricks’ R&D teams and local engineering hubs, helping customers including ABN AMRO, Adidas, Danone, Heineken, HelloFresh, Mercedes-Benz and Nationwide scale AI adoption with  Lakebase, Genie and Unity Gateway. Alongside the investment, Databricks also plans to train nearly 500,000 students and professionals in data and AI skills across EMEA by 2030, and grow its local engineering hubs where core parts of the Databricks Data + AI Platform are  built.\n\n…\n\n**Databricks** announced the second generation of Genie, its AI agent, which is now an always-on, autonomous agent for enterprise work. Employees can give Genie a goal and trust it to keep work moving. Genie takes ownership of the work, delegates to subagents when needed, tracks progress, follows up on missing inputs, and reaches out when a person needs to make a decision. A new inbox highlights what needs attention, and Genie emails users when they are away. Read more about Genie on the Databricks [blog](https://www.databricks.com/blog/new-genie-autonomous-ai-complex-enterprise-work).\n\n…\n\nAI data layer company **Denodo** has signed a multi-year strategic collaboration agreement (SCA) with Amazon Web Services (AWS). The agreement is focused on helping enterprises develop and scale agentic AI on a foundation of real-time, governed data across hybrid and multi-cloud environments. The SCA builds on the existing relationship between AWS and Denodo and marks a new phase of investment to accelerate adoption of the Denodo Platform on AWS worldwide\n\n…\n\nEDB welcomed the general availability of PostgreSQL 19, released in the 30th year of the PostgreSQL open source project. This year’s release delivers long-requested advances in online maintenance and query plan stability. It also reflects a deliberate choice by the PostgreSQL community to ship only the features that earned production trust, at the moment Postgres is becoming the de facto data foundation for AI and agentic applications. The complete release feature list is in the PostgreSQL 19 [__release notes__](https://www.postgresql.org/docs/19/release-19.html).\n\n…\n\n**Hitachi Vantara** has expanded its VSP One platform, enabling businesses to analyse supported data directly within its object storage, reducing the need to move  or copy it into separate systems. The company reports up to 5x faster analytical queries using 80% fewer CPUs. The update also introduces a smaller, four-node starting configuration and new Hitachi iQ Studio capabilities that connect AI applications and agents to governed enterprise data.\n\nThe VSP One Object update provides a more practical path to deploy and scale object storage for AI and analytics workloads, bring data lakehouse analytics closer to where data resides and expand governed access to enterprise data through Hitachi iQ Studio.It also enables organizations to query supported open-format data, including Iceberg, Parquet, JSON and CSV, directly on VSP One Object\n\n…\n\nAP reports the US racketeering conspiracy trial of **Huawei Technologies** approached its end Tuesday, Nov 17, after a prosecutor told jurors that the Chinese telecommunications giant helped Iran spy on its citizens and steal from U.S. companies and a defense lawyer accused the U.S. of embracing a false narrative. Defense attorney Brian Heberlig said in a closing argument in Brooklyn federal court after a two-month-long trial that federal prosecutors “decided Huawei was a criminal enterprise” and built a case from that premise.\n\n…\n\nCloud backup supplier IDrive announced IDrive AI Agents Backup, a new capability that protects the locally stored data behind today's most popular AI coding agents. AI Agents Backup automatically detects and backs up conversations, settings, configurations, CLI data and other supported files from Claude, Cursor, Codex and Cline to the user's IDrive account, and restores them to the same or a new computer in a few steps. Find out more [here.](https://www.idrive.com)\n\n…\n\nSemiconductor material and process supplier Entegris and **Kioxia** announced an expanded partnership this morning to develop materials for future generations of Kioxia's  BiCS FLASH 3D NAND. The two companies are using AI, computational modeling, and digital twins to evaluate a much wider range of candidates virtually. Only the most promising ones move on to physical testing and device validation.  Kioxia’s Geino Technology Center in Mie, Japan, will be the hub, linking Kioxia’s device requirements directly to Entegris’ materials work.\n\n…\n\nAI database technology startup LanceDB has raised $57.5 million in early-stage funding to bring a multimodal lakehouse to researchers, model builders and platform teams. AI is making it easier to create and interpret increasingly complex forms of data, but creating it also brings an infrastructure problem. LanceDB is tackling this problem by providing a unified multimodal data lakehouse, giving researchers, enterprise outfits and AI agents a unified way to access, find, prepare and train on that data without spending weeks managing the underlying infrastructure.\n\nLanceDB’s technology is built on Lance, an open-source data format designed for AI workloads that combines structured information with images, video and other large files. Unlike traditional analytics formats optimized for processing entire tables, Lance supports the rapid retrieval and modification of individual records, making it particularly useful for searching and preparing massive training datasets. Read more [here](https://www.lancedb.com/blog/lancedb-series-b-funding).\n\n…\n\n**Lenovo** revenues for its Q2 jumped 48% to a record $30.3 billion with $1.4 billion in net income. Intelligent Devices Group revenue rose 15% to $17.4 billion. Infrastructure Solutions Group (servers, storage, etc) revenues rose 178% to $11.3 billion, with ISG being called Lenovo’s engine of hyper growth. Solutions & Services Group revenue increased 20% to $2.9 billion.\n\nIn the first full quarter as part of the Lenovo Group, Infinidat delivered its best quarter ever, with the high-end storage portfolio expanding ISG’s enterprise storage capabilities. Lenovo [climbed 43 places](https://cts.businesswire.com/ct/CT?id=smartlink&url=https%3A%2F%2Fnews.lenovo.com%2Fpressroom%2Fpress-releases%2Flenovo-jumps-to-153-on-fortune-global-500-as-ai-driven-business-fuels-growth%2F&esheet=54643970&newsitemid=20261118600654&lan=en-US&anchor=climbed+43+places&index=1&md5=33c524c2bb00fcaf2976df593169a4e1) to #153 on Fortune’s annual [Global 500 List](https://cts.businesswire.com/ct/CT?id=smartlink&url=https%3A%2F%2Ffortune.com%2Franking%2Fglobal500%2F&esheet=54643970&newsitemid=20261118600654&lan=en-US&anchor=Global+500+List&index=2&md5=ecc3194e4aa851a717c0b5f7739cc8cd), making it the highest ranking in the company’s history and one of the most significant advances in recent years.\n\n…\n\n**Liqid**, which supplies composable SW-defined memory and GPU pooling infrastructure, today announced Liqid as a Service, a fully managed, single-tenant AI infrastructure service for workloads that have outgrown a single server. Each Liqid as a Service node starts at 20 GPUs plus 10 TB of CXL memory but can scale to up to 30 GPUs and up to 160 TB of CXL memory to one server. Enterprises get room for large models, long context windows, and large datasets on dedicated infrastructure that Liqid operates for them. Liqid's fabric and Liqid Matrix software connect one host to pools of GPUs and CXL memory, so a single server can use far more memory than its own slots hold. In Liqid initial testing, the same GPUs produced 5-7x more tokens per second with this architecture.Learn more [here](https://www.liqid.com/products/liqid-as-a-service).\n\n…\n\n MongoDB [__announced__](https://www.prnewswire.com/news-releases/mongodb-atlas-for-government-achieves-fedramp-class-d-high-certification-adds-support-for-microsoft-azure-government-302944495.html) that MongoDB Atlas for Government has achieved FedRAMP Class D Certification and added support for Microsoft Azure Government, extending its existing availability on Amazon Web Services (AWS) and Google Cloud Platform (GCP). With this expansion, MongoDB Atlas for Government is now the world’s only FedRAMP Class D Certified, multi-cloud operational data platform.\n\n…\n\nCRN [reports](https://www.crn.com/news/storage/2026/netapp-raises-storage-prices-keystone-rates-by-up-to-35-percent) NetApp has raised prices by up to 40 percent on its hard drive-based storage hardware, software and services and increased prices on select NetApp Keystone storage-as-a-service subscription tiers by as much as 35 percent, citing increased component costs and market dynamics.\n\n…\n\nNexsan has partnered with Bacula Systems and Spectra Logic on a Proven Recovery system bringing together Nexsan’s Unity enterprise storage technology, Bacula Enterprise autonomous backup and recovery software, and Spectra Logic’s tape technology in an integrated architecture designed to help customers maintain control of their recovery resources without depending on external vendors or cloud services during a crisis. These layers provide a defense-in-depth approach to backup and recovery. Nexsan's disk-based storage supports operational recovery and immutable data retention, while Spectra Logic's tape technology adds physical isolation. The architecture allows organizations to select the appropriate recovery path based on the severity and nature of an incident.\n\n…\n\nFabless CXL semiconductor tech startup Panmnesia has established Panmnesia America, its U.S. subsidiary, and appointed Dr. Ellis Wilson, formerly of Microsoft, as the subsidiary’s President. Wilson will lead the new subsidiary and deepen Panmnesia’s engagement with customers and partners in the region.\n\n…\n\nAt SC26 (Supercomputing 2026) in Chicago, Panmnesia stepped up its global business. Ellis Wilson, President of the company’s Panmnesia America, held a series of in-person meetings with customers and partners at the event. Panmnesia confirmed demand from customers including hyperscalers and neoclouds. It also held detailed discussions with memory vendors, CPU vendors and server OEMs/ODMs on the CXL appliances it is building for AI data centers. Panmnesia also announced plans to unveil the follow-on silicon to its PCIe 6.4-CXL 3.2 Fusion Switch in 2027. The new chip reflects the needs the company has identified through its work with hyperscalers and other customers.\n\n…\n\nAlex Brightman, co-founder of [**Pathway**](https://pathway.build/), tells us: “Broadcom's VMware pricing has pushed plenty of estates towards Nutanix, and NC2 lets them land in Azure or AWS without re-platforming the workloads. What slows those moves down is rarely the cluster. It's the cloud underneath it: the networks, permissions, capacity and connectivity back to site that NC2 expects, laid out the way it expects, with region, quota and policy blockers found before anything is built rather than halfway through a migration.\n\n“That foundation is repetitive specialist work, and it's exactly where AI on its own falls over: it writes infrastructure that looks right and fails on deployment. We build the NC2 landing zone on Azure and on AWS in about an hour, at a fixed price, in the customer's own account. The cluster itself is then deployed from NC2 as normal. … for a lot of VMware exits the landing zone is the critical path, and it no longer needs to be.”\n\n…\n\nRedis announced an expansion of Redis Iris, its real-time context engine designed to help businesses take AI agents from prototype to production. Iris now brings together 6 capabilities: Redis Intent Router, Redis Agent Playbooks, Redis Agent Memory, Redis Context Gateway, Redis Lens, and Redis Agent Mesh. They cover memory, context management, intelligent routing and agent-to-agent communication, helping developers build agents that can operate reliably without relying on multiple disconnected infrastructure tools. Highlights include:\n\n- Making AI agents more reliable – giving agents access to relevant information and persistent memory so they can make better decisions and learn from previous interactions\n- Reducing the cost of agentic AI - helping businesses cut unnecessary LLM calls and token consumption by reusing existing knowledge and routing requests more efficiently\n- Building trust and transparency - enabling businesses to trace agent decisions, understand where information comes from and monitor performance\n\nRedis Iris is available today. Intent Router, Agent Playbooks, Agent Memory, and Context Gateway are generally available; Lens and Agent Mesh are in beta. To learn more, visit [redis.io/iris](https://redis.io/iris/).\n\n…\n\n[New research](https://zerolabs.rubrik.com/reports/global-threat-landscape-report-2026-ransomware-extortion-economy) from Rubrik’s Zero Labs reveals that just eight ransomware groups are responsible for half of the 17,000 publicly recorded attacks analysed in 2026, despite 100 groups being active. With 44 groups emerging after March, the threat landscape is constantly shifting. Rather than tracking individual gangs, businesses need to prepare for common attack tactics and prioritise recovery. The findings underline why: nine of the ten most active groups were detected in customer backups, while organisations took a median of 20 days to recover from attacks. \n\nOther key findings include:\n\n- 67% of incident reports with a stated extortion tactic involved stolen data, highlighting the need to prepare for data theft as well as disruption.\n- At least 7% of recovery cases involved attackers attempting to access backup management consoles, with criminals now looking to sabotage ability to recover as well as systems.\n- 68% of forensic findings arrived within 48 hours, yet only 32% of organisations recovered within two weeks and 28% took more than six weeks.\n\n…\n\n[The Lustre Collective](https://thelustrecollective.com/) (TLC), the engineering company founded by veteran Lustre developers to advance Lustre upstream and in the open, today announced a strategic partnership with Do IT Now,  the European HPC and AI infrastructure integrator. Together, the two companies will bring TLC's upcoming software-defined Lustre platform to HPC centers, AI factories and research institutions – running on standard servers from any vendor or in the public  cloud, and backed by a complete delivery and support chain from the first design workshop to day-to-day operations.\n\nTLC's software-defined Lustre platform is currently in joint qualification with Do IT Now as its launch partner. Selected customers can take part in early evaluation programs through Do IT Now ahead of general availability, which is planned for the first quarter of 2027.\n\n…", "url": "https://wpnews.pro/news/storage-news-ticker-19-november", "canonical_source": "https://www.blocksandfiles.com/ai-ml/2026/11/19/storage-news-ticker-19-november/5311097", "published_at": "2026-11-19 16:11:34+00:00", "updated_at": "2026-11-19 16:42:25.871395+00:00", "lang": "en", "topics": ["ai-agents", "agent-protocols", "ai-infrastructure", "ai-tools", "large-language-models"], "entities": ["Airbyte", "Model Context Protocol", "Argonaut Labs", "Salvatore Sanfilippo", "Kimi K3.1 Pro", "NVIDIA RTX PRO 6000 Blackwell", "ClickHouse", "Xata"], "also_reported_by": [], "takeaway": "Airbyte's semantic search is exposed through its MCP server, letting Claude, ChatGPT and other MCP-compatible assistants query business data in natural language and join it with Snowflake and BigQuery structured data.", "recommendation": null, "alternates": {"html": "https://wpnews.pro/news/storage-news-ticker-19-november", "markdown": "https://wpnews.pro/news/storage-news-ticker-19-november.md", "text": "https://wpnews.pro/news/storage-news-ticker-19-november.txt", "jsonld": "https://wpnews.pro/news/storage-news-ticker-19-november.jsonld"}}