ZettaLane's MayaNAS on Google Cloud Marketplace: Unifying HPC and Enterprise NAS on Object Storage for AI Workloads

ZettaLane Systems, a Google Cloud Technology Partner, has announced the general availability of MayaNAS on the Google Cloud Marketplace. This offering provides a unified storage solution that integrates both enterprise NAS capabilities (NFS, SMB with Active Directory) and high-performance parallel file systems like Lustre and pNFS directly on Google Cloud Storage. Notably, MayaNAS achieved impressive results in MLPerf Storage v3.0 benchmarks, demonstrating 32.42 GiB/s checkpoint write and 21.20 GiB/s read on a Llama 3 70B workload, utilizing standard Compute Engine instances and Google Cloud Storage. This signifies a robust capability to handle the demanding I/O requirements of large-scale AI training. This development is crucial for organizations that are increasingly adopting AI and machine learning, as it addresses a long-standing challenge: bridging the gap between the specialized storage needs of HPC/AI workloads and the more general-purpose requirements of enterprise file storage. Historically, these two domains have necessitated separate, often complex, storage infrastructures. By unifying them on Google Cloud Storage, MayaNAS simplifies data management, reduces operational overhead, and eliminates the need for data tiering or staging jobs between fast tiers and object storage. This directly impacts data scientists, MLOps engineers, and IT infrastructure teams by providing a more agile and cost-effective approach to managing their diverse data needs. The broader trend this aligns with is the ongoing convergence of traditional enterprise IT with cloud-native and AI-driven architectures. As AI becomes more pervasive, the underlying infrastructure must adapt to support its unique demands for massive datasets and high-throughput processing, while still maintaining the accessibility and governance required by conventional business applications. The move towards software-defined storage solutions that can abstract away the complexities of the underlying hardware and seamlessly integrate with cloud object storage is a clear indicator of this trend. Other solutions, such as IBM Fusion, also aim to simplify data management for modern applications on platforms like Red Hat OpenShift, highlighting a general industry push towards more integrated and container-native data platforms. In practice, this means practitioners can deploy MayaNAS from the Google Cloud Marketplace, leveraging pay-as-you-go pricing and counting eligible purchases towards existing Google Cloud commitments. The ability to scale serving nodes up for a run and down afterward, while the data remains in place within Google Cloud Storage, offers significant cost optimization for intermittent, high-demand AI workloads. Furthermore, the solution supports data residency and sovereignty requirements by keeping the file system, buckets, and encryption keys within the customer's own project. Practitioners should evaluate MayaNAS for scenarios where they need to consolidate diverse file-based workloads, particularly those involving large-scale AI training, genomics, electronic design automation, or media rendering, and seek to leverage the scalability and cost-effectiveness of Google Cloud Storage without sacrificing performance or operational simplicity.