From the Phoenix Computing Model to Cloud-Native and Agentic AI Runtime Systems

Dynamic Membership Architecture and PhoenixFlight

Dynamic Membership Architecture is a research framework for systems in which computational participants dynamically join, leave, migrate, fail, recover, and retire while preserving execution continuity. The work revisits the Phoenix Computing Model and connects its core ideas to modern distributed systems, Kubernetes, cloud-native orchestration, and agentic AI runtime systems.

Dynamic Membership Architecture and PhoenixFlight were originated and authored by Kinshuk Dutta.

Read on Zenodo Explore PhoenixFlight View GitHub Release View Repository
Kinshuk Dutta Author
v1.0.0 Version
June 14, 2026 Published
CC BY 4.0 License
Working paper Publication type

PhoenixFlight: Reference Implementation of Dynamic Membership Architecture

PhoenixFlight is a Dynamic Membership Control Plane for agentic and distributed workloads. It is the reference implementation of Dynamic Membership Architecture, a framework for systems where computational participants dynamically join, leave, migrate, fail, recover, and retire while preserving execution continuity.


PhoenixFlight addresses a critical enterprise problem: modern AI and distributed automation are becoming dynamic, but governance remains static. Agents fail, tools disappear, models change, services scale, and providers become unavailable. As work moves across participants, identity, context, policy, trust, ownership, and audit lineage often fragment.


PhoenixFlight solves this through governed handoff for dynamic work execution. It treats work as a portable FlightPacket and participants as dynamic members, enabling policy-governed assignment, trust-weighted capability routing, context-preserving handoff, retirement drain, reassignment, and structured audit lineage.

Research Overview

This paper introduces Dynamic Membership Architecture as a generalized framework for execution environments where participants are not static. Resources, services, agents, tools, and computational actors may enter, leave, migrate, fail, recover, or retire during runtime. The framework provides a common vocabulary for analyzing these systems across historical distributed computing, cloud-native infrastructure, and modern agentic AI.

Why It Matters

Distributed Systems Continuity

DMA focuses on preserving logical execution continuity even when the underlying participants change.

Cloud-Native Runtime Design

The framework maps naturally to orchestration patterns found in Kubernetes, scheduling systems, service discovery, workload migration, and runtime control planes.

Agentic AI Governance

DMA extends dynamic membership thinking to autonomous agents, where task handoff, capability discovery, trust, retirement, and auditability become first-class runtime concerns.

Core DMA Primitives

1. Identity

Logical participants remain stable even when physical execution resources change.

2. Discovery

Participants and capabilities must be discoverable at runtime.

3. Assignment

Workloads or goals are assigned based on capability, availability, trust, and policy.

4. Migration

State, context, or responsibility can move between participants.

5. Retirement

Participants can exit safely without breaking execution continuity.

6. Governance

Participation is constrained by policy, authorization, trust, and auditability.

From Architecture to Runtime

Dynamic Membership Architecture

Conceptual framework

PhoenixFlight

Reference implementation

PhoenixFile

Declarative runtime contract

FlightPacket

Portable governed unit of work

Dynamic Member

Agent, service, tool, human, model, API, simulator, optimizer, or provider adapter

Governed Handoff

Policy-controlled transfer of work preserving context and audit lineage

Audit Lineage

Structured record of registration, assignment, handoff, denial, retirement, and completion events

Open-Source Runtime: PhoenixFlight

Dynamic Membership Architecture isn't just a theoretical model. It has been implemented in PhoenixFlight, an open-source runtime environment demonstrating DMA primitives in action.


PhoenixFlight Website Interactive Documentation

Frequently Asked Questions

What is Dynamic Membership Architecture?

Dynamic Membership Architecture is a framework for systems where computational participants dynamically join, leave, migrate, fail, recover, and retire while preserving execution continuity.

What is PhoenixFlight?

PhoenixFlight is a Dynamic Membership Control Plane for agentic and distributed workloads and the reference implementation of Dynamic Membership Architecture.

Who created PhoenixFlight and Dynamic Membership Architecture?

PhoenixFlight and Dynamic Membership Architecture were originated and authored by Kinshuk Dutta.

What problem does PhoenixFlight solve?

PhoenixFlight solves the loss of governance continuity when work moves across dynamic participants. It preserves identity, context, policy, trust, ownership, and audit lineage.

What is governed handoff?

Governed handoff means work can move from one participant to another only when capability, trust, namespace, compliance, policy, and load conditions are satisfied, while preserving context and audit lineage.

What is a FlightPacket?

A FlightPacket is a portable unit of work that carries workload identity, required capabilities, context or checkpoint state, policy tags, ownership history, and audit lineage.

Is PhoenixFlight a replacement for Docker or Kubernetes?

No. Docker packages processes and Kubernetes schedules compute. PhoenixFlight governs ownership, handoff, retirement, and audit of work above the infrastructure layer.

How does PhoenixFlight relate to agent frameworks?

Agent frameworks compose agents and workflows. PhoenixFlight governs dynamic membership, policy, handoff, retirement, and audit continuity for agentic work execution.

How does PhoenixFlight relate to hybrid quantum-classical workloads?

PhoenixFlight does not execute quantum circuits directly or replace quantum providers. It can orchestrate the classical control plane around hybrid quantum-classical workloads by modeling QPUs, simulators, provider adapters, optimization agents, and result auditors as dynamic members.

Citation

Dutta, K. (2026). Dynamic Membership Architecture: From the Phoenix Computing Model to Cloud-Native and Agentic AI Runtime Systems (v1.0.0). Zenodo. https://doi.org/10.5281/zenodo.20693483
@misc{dutta2026dynamicmembershiparchitecture, author = {Dutta, Kinshuk}, title = {Dynamic Membership Architecture: From the Phoenix Computing Model to Cloud-Native and Agentic AI Runtime Systems}, year = {2026}, publisher = {Zenodo}, version = {v1.0.0}, doi = {10.5281/zenodo.20693483}, url = {https://doi.org/10.5281/zenodo.20693483} }