Skip to content

roadmap.rst

Latest commit

 

History

History
86 lines (55 loc) · 3.27 KB

roadmap.rst

File metadata and controls

86 lines (55 loc) · 3.27 KB

NVIDIA FLARE Roadmap

This page outlines planned features and target release milestones for upcoming NVIDIA FLARE versions. Dates and features are subject to change.

Note

This roadmap reflects current planning and is provided for informational purposes. Feature scope and release timing may shift as development progresses.

FLARE 2.8.0 — Target: Q2 2026

Native Kubernetes Support

  • Separate the parent control pod from the job execution pod, enabling independent lifecycle management and better resource isolation
  • Simplified deployment across major cloud Kubernetes environments (GKE, EKS, AKS, and on-prem)

Improved Docker Deployment

  • Separate parent container from job execution container, mirroring the Kubernetes pod separation model
  • Ready-to-use Dockerfiles provided for common deployment scenarios, reducing setup friction

Multi-Study Support

  • Enable multiple concurrent studies within a single FLARE deployment
  • Enforce data isolation between studies via Docker and Kubernetes pod-level data separation

Distributed Provisioning

  • Enable the distributed provisioning workflow so site administrators can generate their own key pairs locally and receive signed certificates from the project administrator
  • Eliminates the need for centralized private key generation and distribution

Expanded CLI Commands

  • Extend the nvflare CLI to cover all FLARE Admin Console commands
  • Enables full administrative control from the command line without requiring the interactive console

Server-Side Memory Optimization

  • Reduce server-side memory usage during federated learning jobs
  • Improved memory management for large model and large dataset workloads

FLARE 2.9.0 — Target: Q3 2026

New Collab API

  • Introduce a new Collaboration (Collab) API designed to improve research productivity
  • Enables more flexible and composable FL workflow definitions with reduced boilerplate

FLARE Agent Readiness

  • Platform features enabling FLARE to be used as a backend for AI agent workflows

Better Kubernetes User Experience

  • Simplified Kubernetes deployment and operational experience building on the 2.8.0 foundation
  • Usability improvements for data scientists and operators managing multi-site Kubernetes deployments

Slurm Support

  • Better integration with Slurm workload managers for HPC cluster environments
  • Enable FL jobs to run natively within Slurm-managed compute environments

FLARE 2.10.0 — Target: Q4 2026

Advanced Kubernetes Enhancements

  • Advanced Kubernetes feature set building on prior releases
  • Deeper platform integration and operational controls for large-scale multi-site Kubernetes deployments

Confidential Federated AI Support

Building on existing support for AMD SEV-SNP CPU CVMs with NVIDIA GPUs and Azure Confidential Computing:

  • Intel TDX CPU support for CPU-based confidential computing workloads
  • CoCo (Confidential Containers) support for container-level confidential execution
  • Expanded Cloud Service Provider (CSP) integration beyond Azure to additional major cloud platforms