Parallelstore

High performance, managed parallel file service

Parallelstore is based on Intel DAOS and delivers up to 6x greater read throughput performance compared to competitive Lustre scratch offerings. 

Features

Fast, scalable performance

High bandwidth, high IOPS, and ultra-low latency, utilizing byte-addressable media to store metadata and small I/O, and locally attached NVMe with software-managed redundancy for bulk I/O.

Accelerate time to value

Configure Parallelstore to ideally fit your use case and build a system of the right scale for extreme generative AI and HPC simulation use cases.

Futureproof your architecture

Building with Parallelstore prepares your business for HPC scale, AI/ML convergence, and Kubernetes integration, which means you can grow and scale with minimal disruption. Distributed metadata management, extreme IOPS, and key value store architecture are well-aligned to emerging patterns in AI workloads. 

Open source flexibility

Parallelstore is built on Intel DAOS, which is open source. We have a long history of leadership in open source—from projects like Kubernetes, TensorFlow, and more. Open source gives you the flexibility to deploy—and, if necessary, migrate—critical workloads across or off public cloud platforms. Bring products and services to market faster, without operational overhead or requirement of specialized skills.

How It Works

"With DAOS on Google Cloud, users can easily and quickly provision storage clusters that can scale to similar performance levels as similar on-premises hardware, but then also be able to dynamically grow and shrink those clusters as needed." - Andrey Kudryavtsev, Intel

DAOS whitepaper image

Common Uses

AI/ML training

For systems that need high-speed access to very large files or millions of small files, parallel file systems can outperform NFS and object storage. The latency offered by parallel file systems is significantly lower than other options, which can affect the maximum IOPS, making Parallelstore a great choice for AI/ML workload scratch space.
Explore details and tutorials

    For systems that need high-speed access to very large files or millions of small files, parallel file systems can outperform NFS and object storage. The latency offered by parallel file systems is significantly lower than other options, which can affect the maximum IOPS, making Parallelstore a great choice for AI/ML workload scratch space.
    Explore details and tutorials

      Quantitative trading analysis

      Parallelstore is ideally suited to provide scratch analysis of intermediate data in high performance quantitative analysis and trading, and is also well-suited to other complex, high-speed financial services use cases such as fraud detection.
      Explore documentation

        Parallelstore is ideally suited to provide scratch analysis of intermediate data in high performance quantitative analysis and trading, and is also well-suited to other complex, high-speed financial services use cases such as fraud detection.
        Explore documentation

          Computer-Aided Engineering (CAE)

          Parallelstore is a good option for manufacturing, auto, aerospace, and biomedical applications doing computational fluid dynamics (CFD), complex modeling, crash simulation, chemical engineering, and more.
          Explore documentation

            Parallelstore is a good option for manufacturing, auto, aerospace, and biomedical applications doing computational fluid dynamics (CFD), complex modeling, crash simulation, chemical engineering, and more.
            Explore documentation

              Pricing

              How Parallelstore pricing worksParallelstore pricing is based on instance capacity and the location where your Parallelstore instance is provisioned.
              ProductPricing

              Instance Pricing

              From $0.0001917808 per GiB per hour

              Data transfer pricing using the Parallelstore API

              From $0.001 per GiB transferred

              How Parallelstore pricing works

              Parallelstore pricing is based on instance capacity and the location where your Parallelstore instance is provisioned.

              Instance Pricing

              Pricing

              From $0.0001917808 per GiB per hour

              Data transfer pricing using the Parallelstore API

              Pricing

              From $0.001 per GiB transferred

              Pricing calculator

              Estimate your monthly charges using Google Cloud.

              Custom quote

              Connect with our sales team to get a custom quote for your organization.

              Begin testing now

              Try Parallelstore today

              Understand the technical benefits of DAOS

              Learn about parallel file systems for HPC

              When to use parallel file systems

              Google Cloud
              • ‪English‬
              • ‪Deutsch‬
              • ‪Español‬
              • ‪Español (Latinoamérica)‬
              • ‪Français‬
              • ‪Indonesia‬
              • ‪Italiano‬
              • ‪Português (Brasil)‬
              • ‪简体中文‬
              • ‪繁體中文‬
              • ‪日本語‬
              • ‪한국어‬
              Console
              Google Cloud