Oakestra: Edge Computing Orchestration Framework

Oakestra is an open-source, lightweight hierarchical orchestration framework designed specifically for edge computing environments. Co-founded by Nitinder Mohan, Oakestra addresses the unique challenges of managing applications across distributed, heterogeneous edge infrastructures.

Overview

As edge computing becomes increasingly important for latency-sensitive applications like IoT, autonomous vehicles, and AR/VR, traditional cloud orchestration tools fall short. Oakestra fills this gap by providing:

Hierarchical Architecture: Multi-tier management spanning cloud, edge, and far-edge resources
Lightweight Design: Minimal resource footprint suitable for constrained edge devices
Heterogeneous Support: Unified management of containers, VMs, and unikernels
Geographic Awareness: Location-based service placement and routing
Network Function Virtualization: Built-in support for service chains and network functions

Key Challenges Addressed

1. Resource Heterogeneity

Edge environments comprise diverse devices with varying capabilities. Oakestra provides a unified interface for managing:

High-power edge servers
Mid-tier gateways
Resource-constrained IoT devices
Cloud resources for overflow and backup

2. Geographic Distribution

Unlike centralized cloud data centers, edge resources are geographically distributed. Oakestra enables:

Location-aware application placement
Dynamic service migration based on user mobility
Multi-site application deployment with geographic constraints

3. Network Constraints

Edge networks have limited and variable bandwidth. Oakestra optimizes for:

Reduced control plane overhead
Efficient state synchronization
Adaptive routing based on network conditions

4. Application Requirements

Modern edge applications have diverse needs. Oakestra supports:

Ultra-low latency requirements (< 10ms)
Real-time data processing
Privacy-sensitive workloads (local processing)
High availability through multi-site deployment

Architecture

Oakestra implements a three-tier hierarchical architecture:

Cloud Tier

Global orchestrator managing multiple clusters
Centralized monitoring and analytics
Policy enforcement and governance
Integration with cloud services

Edge Tier

Cluster orchestrators managing edge sites
Local scheduling and placement decisions
Service mesh coordination
Edge-to-edge communication

Far-Edge Tier

Worker nodes running applications
Lightweight agents with minimal overhead
Support for various virtualization technologies
Local health monitoring

Technical Innovations

Hybrid Virtualization

Oakestra supports both containers and virtual machines, enabling:

Legacy application migration to the edge
Security isolation for sensitive workloads
Hardware-specific virtualization features
Flexible deployment models

Publication: Supporting Hybrid Virtualization Orchestration for Edge Computing (EdgeSys 2025, Best Paper Award)

On-Demand Container Partitioning

Dynamic partitioning of applications for distributed execution:

Split ML models across edge and cloud
Reduce data movement and latency
Optimize resource utilization
Enable inference on constrained devices

Publication: On-Demand Container Partitioning for Distributed Machine Learning (USENIX ATC 2025)

Geographic Checkpointing

Location-aware service placement using geographic constraints:

Route traffic to nearest service instance
Maintain service replicas within regions
Support for mobility-aware applications
Compliance with data locality requirements

Use Cases

IoT Application Management

Deploy and manage thousands of IoT applications across distributed edge infrastructure with automatic placement based on sensor locations.

Augmented Reality

Low-latency AR/VR processing at the edge with dynamic offloading to nearby edge servers based on user movement and compute requirements.

Autonomous Vehicles

Real-time data processing for vehicle-to-everything (V2X) communications with geographic service placement along roadways.

Smart Cities

Coordinated deployment of city services across municipal edge infrastructure with hierarchical management and policy enforcement.

Key Results

Performance

Deployment Speed: 3x faster than Kubernetes for edge deployments
Resource Overhead: 50% lower memory footprint compared to K3s
Scaling: Tested with 1000+ edge nodes across multiple sites
Latency: Sub-10ms service-to-service communication

Recognition

USENIX ATC 2023: Oakestra paper accepted (18.4% acceptance rate)
Artifact Badges: Available, Functional, Reproduced
Community Adoption: Growing user base across academia and industry

Open Source

Oakestra is fully open source and available on GitHub:

Repository: github.com/oakestra/oakestra
Website: oakestra.io
Documentation: Comprehensive guides and API references
Community: Active Discord and mailing list

Contributing

We welcome contributions from the community:

Feature development and bug fixes
Documentation improvements
Use case studies and benchmarks
Integration with other tools and platforms

Current Development

Active areas of development include:

Service Mesh Integration: Advanced traffic management and observability
AI/ML Workload Support: Optimized scheduling for training and inference
Network Function Support: Enhanced NFV capabilities
Energy Optimization: Power-aware scheduling and placement
Security Enhancements: Zero-trust networking and attestation

Research Collaborations

Oakestra is being used in several research projects:

EDGELESS: EU Horizon 2020 project on cognitive edge computing (€5.4M)
6G Research: Next-generation network infrastructure studies
Smart City Deployments: Real-world edge infrastructure management
Academic Partnerships: Multiple universities using Oakestra for research

Team

Core Contributors

Nitinder Mohan - Co-founder, Architecture Lead
Giovanni Bartolomeo - Lead Developer
Jörg Ott - Research Advisor

Collaborators

Technical University of Munich (TUM)
Delft University of Technology (TU Delft)
Community contributors worldwide

Publications

Oakestra: A Lightweight Hierarchical Orchestration Framework for Edge Computing Giovanni Bartolomeo, Mehmet Mert Bese, Nitinder Mohan, Jörg Ott USENIX ATC 2023
Supporting Hybrid Virtualization Orchestration for Edge Computing Giovanni Bartolomeo, Patrick Sabanic, Nitinder Mohan, Jörg Ott EdgeSys 2025 (Best Paper Award)
On-Demand Container Partitioning for Distributed Machine Learning Giovanni Bartolomeo, Navidreza Asadi, Wolfgang Kellerer, Jörg Ott, Nitinder Mohan USENIX ATC 2025
Oakestra: An Orchestration Framework for Edge Computing (Demo) Giovanni Bartolomeo, Simon Bäurle, Nitinder Mohan, Jörg Ott ACM SIGCOMM 2022

Get Started

Quick Start

# Install Oakestra
curl -sfL https://get.oakestra.io | sh -

# Deploy your first application
oakestra app deploy examples/hello-edge

# Monitor deployment
oakestra app list

Learn More

Contact

Interested in using or contributing to Oakestra? Get in touch:

GitHub Issues: For bugs and feature requests
Discord: Join our community chat
Email: n.mohan@tudelft.nl for research collaborations

Join us in building the future of edge computing orchestration! 🚀

Oakestra - Edge Computing Orchestration Framework

Project Team