Hosted lakeFS Enterprise (lakeFS Cloud)¶

lakeFS Cloud is a single-tenant, fully managed lakeFS service that provides high availability, auto-scaling, support, and production-grade features. It gives you the benefits of lakeFS without any of the work of installing, upgrading, and scaling it yourself, so there is no infrastructure to run and no maintenance to plan. Treeverse manages every lakeFS service on your behalf, including managed garbage collection, performs upgrades transparently, and backs the service with an uptime SLA.

Security¶

lakeFS Cloud is SOC2 Type II compliant, and you can contact us to obtain the certification. It version-controls your data without accessing it by using presigned URLs, which makes it both efficient and secure. On top of that you get the full Enterprise security package: Role-Based Access Control, Single Sign-On with SAML, OIDC, AD FS, Okta, and Azure AD, SCIM provisioning, short-lived token authentication, AWS IAM role authentication, auditing, and Private Link so that access is only allowed from your own cloud accounts.

Additional functionality¶

Beyond a managed and secure way to run lakeFS, the hosted edition includes lakeFS Mount for exposing repository data as a local filesystem, the Iceberg REST Catalog for managing Iceberg tables alongside other data, Metadata Search for querying versioned objects by their metadata, lakeFS for Snowflake for full Snowflake integration including managed Iceberg tables, lakeFS Cross-Cloud for centrally managing repositories spanning multiple clouds and on-prem, and transactional mirroring for replicating repositories into consistent read-only copies in remote locations.

How lakeFS Cloud interacts with your infrastructure¶

Treeverse hosts and manages a dedicated lakeFS instance that interfaces with data held in your own object store, such as S3.

flowchart TD
    U[Users] --> LFC

    subgraph Your Infrastructure
    IAMM[lakeFS Managed GC IAM Role] --> ObjectStore[Client's Object Store]
    IAMA[lakeFS Application IAM Role] --> ObjectStore
    end

    subgraph Treeverse's Infrastructure
    MGC[Managed Garbage Collection] --> EMR[Elastic Map Reduce]
    EMR --> IAMM
    MGC --> CP
    CP[Control Plane]
    LFC --> CP

        subgraph Client's Tenant
        LFC[lakeFS Cloud] --> DB[Refstore Database]
        end

    LFC --> IAMC[lakeFS Connector IAM Role]
    IAMC -->|ExternalID| IAMA
    end

lakeFS Cloud manages a dedicated instance that reads and writes data directly in your object store

Setting up lakeFS Cloud¶

On AWS and Azure, follow the self-service setup wizard at lakeFS Cloud. On GCP, contact us for onboarding instructions.

Scalability model¶

By default, a lakeFS Cloud installation supports 1,500 read operations per second and 1,500 write operations per second across all branches on all repositories within a region. This limit can be increased by contacting support.

Each lakeFS branch can sustain up to 1,000 write operations per second and 3,000 read operations per second, and this scales horizontally. For example, with 10 concurrent branches a repository could sustain 10k writes per second and 30k reads per second, assuming load is distributed evenly between them. Reading committed data, such as from a commit ID or tag, can be scaled horizontally to any desired capacity, and defaults to roughly 5,000 reads per second.