lakeFS Cloud

lakeFS Cloud is a fully-managed lakeFS solution provided by Treeverse, implemented using our best practices, providing high availability, auto-scaling, support and enterprise-ready features.

lakeFS Cloud Features

• Role-Based Access Control
• Auditing
• Single-Sign-On (including support for SAML, OIDC, AD FS, Okta, and Azure AD)
• Managed Garbage Collection
• Private-Link
• Transactional Mirroring
• SOC 2 Type II Compliance

How lakeFS Cloud interacts with your infrastructure

Treeverse hosts and manages a dedicated lakeFS instance that interfaces with data held in your 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

Setting up lakeFS Cloud

AWS / Azure

Please follow the self-service setup wizard on lakeFS Cloud

GCP

Please contact us for onboarding instructions.

Scalability Model

By default, a lakeFS Cloud installation supports:

• 1,500 read operations/second across all branches on all repositories within a region
• 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 a maximum of 1,000 write operations/second and 3,000 read operations per second. This scales horizontally, so for example, with 10 concurrent branches, a repository could sustain 10k writes/second and 30k reads/second, assuming load is distributed evenly between them.

Reading committed data (e.g. from a commit ID or tag) could be scaled up horizontally to any desired capacity, and defaults to ~5,000 reads/second.