Keeping pace with news and developments in the real-time analytics and AI market can be a daunting task. Fortunately, we have you covered with a summary of the items our staff comes across each week. And if you prefer it in your inbox, sign up here!
MLCommons announced the results of its MLPerf Storage v3.0 benchmark suite, which measures the performance of storage systems for machine learning (ML) workloads in an architecture-neutral, representative, and reproducible manner. Version 3.0 expands the tests in the suite to represent the breadth of storage workloads that AI systems can generate and adds support for an S3 object storage access layer alongside the existing POSIX layer.
The suite includes two new benchmarks.
Version 3.0 adds a new KV Cache test, which measures storage performance for LLM inference cache read/write operations. KV caching is a widely adopted technique to increase performance in transformer-based AI inference applications, particularly autoregressive ones such as LLMs that iteratively access the same key-value vectors.
It also adds a new Vector Database (VDB) test that measures storage performance for vector indexing and querying workloads. Vector databases are used to store high-dimensional data that exceed the limits of traditional database architectures, and they are frequently used by AI applications to store unstructured data such as text and media.
In addition, Version 3.0 adds new support for an S3 object storage access layer, alongside and as an alternative to the existing POSIX-compliant layer. Adding this capability in v3.0 enables performance comparisons across a wider range of hosted storage options for the same workload.
A Peek at the Results
Nineteen organizations submitted to this round of the benchmark. On-premises submissions for the checkpointing write test achieved a median rate of 14 GB/second per watt, with a maximum of 201. Similarly, submissions for the UNet3D read test achieved a median of 34 GB/second per watt, with a maximum of 277.
To view the detailed results for MLPerf Storage v3.0, visit Storage benchmark results.
Real-time analytics news in brief
Amazon Web Services (AWS) announced the general availability of its next-generation Amazon Elastic Compute Cloud (Amazon EC2) R9g and R9gd instances. Powered by the new AWS Graviton5 processor, these memory-optimized instances represent a significant architectural leap forward for enterprise cloud infrastructure. Ideal workloads include in-memory databases and caches (Valkey, Redis OSS, Memcached), real-time analytics, Kubernetes and containerized applications, and distributed data processing.
Boomi announced major platform innovations designed to solve critical barriers to enterprise AI adoption. At the core of these updates is Boomi’s Agent Control Plane, AI-native infrastructure that securely connects AI agents to core business systems, provides governance over agent activity, and controls runaway AI costs. This infrastructure runs flexibly across public cloud, the customer’s own cloud (VPC), or on-premises, directly supporting data and digital sovereignty, and giving organizations greater operational control over their AI estate.
CIQ announced Fuzzball 4.2, a release of its turnkey sovereign AI and HPC orchestration platform that helps organizations turn training and inference jobs from fragile infrastructure projects into reliable production workloads. The solution is agent-ready infrastructure. A new MCP server lets AI agents inspect a Fuzzball environment directly and draft, submit, and monitor workflows.
IBM rolled out a wave of new features within its AI coding platform, IBM Bob, that center on a more controlled and secure developer experience. Overall, IBM Bob can now run in editors already in use, enable loop engineering with lifecycle hooks, and give admins better central policy enforcement and audit capabilities within existing security platforms. Some new features include Native Agent Client Protocol (ACP) support, centralized group policies and audit controls, and more.
The Institute of Foundation Models (IFM) introduced K2 Horizon, a new fleet of six AI foundation models ranging from 0.9 billion to 375 billion parameters. The new models are fully open, including model weights, code, training data, and methodologies, allowing researchers and developers to inspect, reproduce, and adapt the models for their own work.
KBC (A Yokogawa Company) announced the launch of Petro-SIM 7.7, an advanced process simulation, optimization, and digital twin platform for engineers and safety specialists. Built on nearly 50 years of refinery and process engineering expertise, Petro-SIM 7.7 integrates AI/ML-enabled hybrid modeling with first-principles simulation within one platform. By combining engineering physics with machine learning, it helps engineers make safer, more informed decisions while maintaining high-fidelity digital twins and optimizing large-scale operations across traditional and emerging energy systems.
OpenMatter Network announced a significant expansion of its platform with new capabilities that make it easier for enterprises to build, deploy, and collaborate using sensitive data and AI while maintaining cryptographic control over how information is accessed, computed, and shared. The new capabilities, available now as part of the commercially available OpenMatter Network platform, span secure application development, AI model management, privacy-preserving machine learning, and data collaboration.
Syspro launched Syspro Torque, an industrial AI platform that detects operational problems, recommends the next step, and takes approved action inside the systems manufacturers already run. Torque is built to work with any ERP. It connects to any mix of technology on the floor, whether the operation runs in the cloud, on-premises, or both. Additionally, Torque connects to legacy SCADA systems, MES platforms, warehouse hardware, and other sources through MCP connectors, with no bespoke middleware required.
Think announced Think Fabric, a new category of unified AI infrastructure that combines intelligent software with high-performance hardware. Think Fabric is designed to address some of the most significant efficiency bottlenecks in AI infrastructure today, including cooling, density, GPU utilization, and infrastructure cost, enabling enterprises, governments, and AI-native organizations to deploy high-performance AI with greater efficiency, flexibility, and control.
Partnerships, collaborations, and more
Coder announced Coder Agent Relay, a self-hosted execution environment for cloud coding agents, with SpaceXAI as its launch partner. Cursor Cloud Agents can now run inside Coder workspaces on infrastructure the customer already operates. Developers keep the Cursor experience they know (e.g., app, web, and mobile) while Cursor continues to run the agent loop, including inference and planning. Tool calls execute in Coder environments on the customer’s network, so source code, secrets, and internal services stay on machines they control.
Commvault announced a new integration with CrowdStrike that makes Commvault cyber recovery actions available as native steps within Charlotte Agentic SOAR workflows. Specifically, the new purpose-built connector enables security teams to incorporate Commvault cyber recovery actions directly into workflows orchestrated by Charlotte Agentic SOAR and rapidly accelerate investigation, response, and recovery.
DE-CIX announced the implementation of georedundant connectivity to Microsoft Azure in cities spanning North America, Europe, and Asia. Azure ExpressRoute Metro connectivity solution involves dual-homed connections to two distinct ExpressRoute locations within the same city, offering very high availability for sensitive AI workloads and critical resources in Microsoft Azure. The highly resilient architecture is now available at DE-CIX’s carrier- and data-center-neutral Cloud and AI Exchanges in New York, Frankfurt, Madrid, Amsterdam, and Singapore, with more locations to follow.
DeepInfra, which offers a purpose-built cloud platform for high-throughput AI inference, announced a partnership with humans& for adeployment of DeepInfra’s DeepCluster offering. The new cluster, located in Hillsboro, Ore., features more than a thousand NVIDIA B300 GPUs and more than two megawatts of power capacity, providing the large-scale infrastructure required to support frontier AI model development.
IBM and Confluent announced the launch of IBM Granite Time Series models in Early Access on Confluent Cloud, bringing forecasting and anomaly detection directly to enterprise data streams. The integration pairs IBM’s time-series foundation models with Confluent’s data streaming platform, so teams can analyze operational data as it’s generated. Teams can call the models from Apache Flink on Confluent Cloud, turning live signals into forecasts, anomaly alerts and downstream actions, all without standing up a separate machine learning environment.
NetApp and Amazon Web Services (AWS) announced that AWS Transform now supports Amazon FSx for NetApp ONTAP, making it easier for enterprises to move applications and data to AWS. The combined solution gives customers the option to migrate block storage as part of the same migration wave that handles compute and network, eliminating the need for intermediate storage platforms, separate migration tools, and the additional cost and risk they introduce.
NVIDIA announced that it has agreed to acquire Hugging Face for approximately $12.9 billion. NVIDIA said Hugging Face will remain an open platform supporting models, frameworks, cloud providers, and computing platforms from across the AI ecosystem. Together, the companies will scale Hugging Face’s platform, strengthen its infrastructure and expand access to AI for developers and institutions worldwide.
Qlik announced expanded availability of its Model Context Protocol (MCP) server with Amazon Web Services (AWS) and Databricks, making it easier for customers to connect assistants and agents to Qlik’s governed data, analytics context, and transformation capabilities in the tools they already use. With availability now in AWS Marketplace and Databricks Marketplace, Qlik is extending that trusted intelligence layer into two of the most important environments where customers are building, deploying, and using AI.
Teradata announced the integration between the Teradata Autonomous Knowledge Platform and Microsoft OneLake, enabling enterprise customers to run Teradata’s high-performance enterprise AI directly on data stored in OneLake, without extract, transform, and load (ETL) pipelines, data duplication, or migration. Built on open Apache Iceberg standards, the integration allows Teradata users to query OneLake tables in place using standard Iceberg APIs, with cross-platform authentication and access controls handled natively.
VAST Data and CrowdStrike announced new integrations that will bring enterprise-grade cybersecurity across the infrastructure, data, and AI workloads supporting production AI. To that end, VAST’s AI infrastructure platform natively supports the CrowdStrike Falcon sensor, establishing a foundation for deeper integrations with Falcon Next-Gen SIEM and Falcon Guardian, CrowdStrike’s new AI Detection and Response (AIDR) solution, across the VAST AI Operating System. Together, VAST and CrowdStrike are integrating security across the systems running AI, access to the proprietary data that fuels it, and the data pipelines preparing that information for AI applications.
Yugabyte announced the new Yugabyte Partner Program, a unified global initiative built to help partners modernize enterprise database workloads and deliver AI-ready applications. The new program provides partners with a consistent framework for engaging with Yugabyte. Through the program, partners can build recurring database modernization and AI infrastructure practices while helping customers replace legacy databases with YugabyteDB, the distributed PostgreSQL database built for resilient, globally distributed, business-critical applications.
If your company has real-time analytics news, send your announcements to ssalamone@rtinsights.com.
In case you missed it, here are our most recent weekly real-time analytics news roundups:
- Real-time Analytics News for the Week Ending August 29
- Real-time Analytics News for the Week Ending August 22
- Real-time Analytics News for the Week Ending August 15
- Real-time Analytics News for the Week Ending August 8
- Real-time Analytics News for the Week Ending August 1
- Real-time Analytics News for the Week Ending July 25
- Real-time Analytics News for the Week Ending July 18
- Real-time Analytics News for the Week Ending July 11
- Real-time Analytics News for the Week Ending July 4
- Real-time Analytics News for the Week Ending June 27
- Real-time Analytics News for the Week Ending June 20
- Real-time Analytics News for the Week Ending June 13
- Real-time Analytics News for the Week Ending June 6
