New Managed Lustre advancements deliver ultra-high throughput, scalable KV-cache innovation, and unified storage to power next-generation AI and HPC workloads in the cloud
DDN, a global leader in AI data platforms, has unveiled a new wave of innovations built around Google Cloud Managed Lustre at the flagship Google Cloud Next 2026. These advancements, powered by DDN’s deep expertise in Lustre file systems and its EXAScaler technology, mark a significant leap forward in enabling large-scale artificial intelligence (AI) and high-performance computing (HPC) workloads in cloud environments.
The announcement highlights a major milestone in cloud infrastructure evolution, as Managed Lustre on Google Cloud now delivers performance scaling up to 10 terabytes per second. This level of throughput, combined with enhanced elasticity and cost efficiency, allows enterprises to run some of the most demanding AI and HPC workloads with unprecedented speed and scalability. The solution is designed to support the entire AI lifecycle—from model training and fine-tuning to inference and large-scale simulations—within a unified, high-performance data platform.
According to Alex Bouzari, the launch represents more than just a product enhancement; it signals a broader transformation in how data infrastructure supports AI at scale. He emphasized that the collaboration with Google Cloud is delivering one of the fastest-growing and highest-performing managed Lustre services in the industry, purpose-built to address the real-world demands of modern AI workloads. The initiative reinforces DDN’s position as a leader in AI data platforms while enabling customers to innovate more rapidly, reduce operational costs, and deploy solutions with greater confidence.
At the technical core of this innovation is a POSIX-compliant parallel file system that provides both high throughput and low latency—two critical requirements for modern AI and HPC applications. Managed Lustre is increasingly being adopted across industries such as financial services, life sciences, robotics, autonomous systems, and advanced scientific research. These sectors rely on massive data pipelines and compute-intensive processes, making efficient storage and data access a foundational requirement.
The platform supports a wide range of use cases, including large-scale training and fine-tuning of large language models (LLMs), high-throughput inference workloads, retrieval-augmented generation (RAG), and key-value (KV) cache acceleration. It also enables complex simulations, financial modeling, machine vision applications, and multimodal AI systems that integrate text, image, and sensor data.
One of the most notable innovations introduced at the event is the use of Managed Lustre as a shared KV-cache layer for AI inference. Traditionally, KV-cache is stored in host memory, which can limit scalability and increase costs due to redundant computations. By leveraging Lustre’s distributed architecture, ultra-low latency, and high aggregate throughput, organizations can now externalize and share KV-cache across clusters. This approach significantly improves both performance and efficiency, enabling near-unlimited shared cache capacity for large-scale inference workloads.
Benchmark testing demonstrates the tangible impact of this innovation. Organizations using Managed Lustre for shared KV-cache have achieved up to a 75% improvement in total inference throughput, along with more than a 40% reduction in mean time to first token compared to traditional in-memory approaches. These gains translate directly into faster, more responsive AI applications and a substantial reduction in the cost of running inference at scale—an increasingly important factor as AI deployments grow in size and complexity.
The collaboration between DDN and Google Cloud is a key enabler of these advancements. DDN contributes decades of expertise in extreme-scale data systems and parallel file system architecture, while Google Cloud provides the underlying infrastructure, including elastic compute resources, global networking, and advanced storage technologies such as Hyperdisk. Additionally, customers benefit from access to specialized AI accelerators, including Tensor Processing Units (TPUs), which further enhance performance for training and inference workloads.
This integrated approach allows organizations to seamlessly scale their AI operations without the traditional constraints of on-premises infrastructure. The cloud-native nature of Managed Lustre ensures that resources can be dynamically provisioned and adjusted based on workload demands, improving both efficiency and cost management.
Real-world adoption is already demonstrating the value of this platform. Sony Honda Mobility, for example, has leveraged Managed Lustre to accelerate AI model training for its AFEELA Intelligent Drive system. According to Motoi Kataoka, the solution enabled a threefold increase in training scalability compared to other Google Cloud storage options. This highlights the platform’s ability to support cutting-edge applications in areas such as autonomous driving and intelligent mobility systems.
Another significant enhancement introduced at Google Cloud Next 2026 is the implementation of a unified storage tier that dynamically manages both hot and cold data. Traditional storage architectures often rely on multiple tiers, leading to complexity, performance inconsistencies, and increased management overhead. The new approach eliminates these challenges by providing a single, dynamic tier that delivers high performance for frequently accessed data while maintaining cost efficiency for less active datasets. This simplification not only improves performance predictability but also reduces operational complexity for enterprise users.
From an industry perspective, the innovations surrounding Managed Lustre are setting new benchmarks for cloud-based AI and HPC infrastructure. The combination of high performance, scalability, and cost efficiency is driving rapid adoption among organizations seeking to modernize their data platforms and support increasingly complex workloads.
Kirill Tropin noted that the partnership with DDN exemplifies the impact of combining deep infrastructure expertise with cloud-scale innovation. By integrating advanced data systems with Google Cloud’s global infrastructure, the collaboration enables customers to run their most demanding workloads with the performance, simplicity, and scalability required for both current and future applications.
As AI continues to evolve, the importance of robust data infrastructure will only increase. Training larger models, processing more diverse data types, and deploying AI at scale all require systems that can handle massive data volumes with minimal latency and maximum efficiency. The DDN–Google Cloud Managed Lustre solution addresses these needs by providing a unified platform that supports the full spectrum of AI and HPC workloads.
In conclusion, the innovations unveiled by DDN and Google Cloud represent a significant step forward in redefining how AI and high-performance computing are executed in the cloud. By combining cutting-edge storage technology, scalable infrastructure, and advanced AI capabilities, the partnership is enabling organizations to push the boundaries of what is possible—accelerating innovation, improving performance, and reducing costs across the AI lifecycle.
Source link: https://www.businesswire.com
