How Akka runs

Every system you build with Akka runs on the Akka runtime. The runtime is the same wherever it runs, so your services are the same code in every deployment option. What changes between options is who operates the runtime and the infrastructure under it.

The runtime

Akka keeps state together with the code that owns it. Instead of stateless services on one side and a database on the other, each component holds its own state in memory and records every change durably as events and snapshots. The application is its own system of record. When a process or node fails, the runtime rebuilds the state of each component from its events, with no operator action.

Services form a cluster with no primary node. State is sharded across the nodes, each entity has one owner at a time, and requests are routed to the owner wherever it runs. As nodes join or leave, shards rebalance. Nodes communicate over encrypted gRPC without a message broker, and split-brain resolution keeps the cluster consistent when the network partitions.

A project can span several regions, clouds and data centers as one system. State is replicated to every region, reads are served locally, and you choose how writes are handled in each region. You can filter what is replicated, and keep data in the regions where it is allowed to live.

See The runtime, Deployment and operations, State and read models and Multi-region operations.

Deployment options

Option Where it runs Who operates it Start here

Serverless

Akka’s cloud, shared infrastructure, in regions on every major hyperscaler

Akka

Deploy to akka.io

Dedicated

Akka’s cloud, a single-tenant region

Akka

How Dedicated works

BYOC

Your cloud account on AWS, Azure or GCP

Akka, on infrastructure you own

Get started with BYOC

BYOK8s

A Kubernetes cluster you provide

Akka operates the platform; you own the cluster

Get started with BYOK8s

Self-managed

Infrastructure you choose: bare metal, VMs, containers or edge

You

Self-managed operations

Serverless is pay as you go, paid by credit card. You create an account at console.akka.io and deploy from the Akka CLI.

In BYOC and BYOK8s, application data stays in your environment. A limited set of platform interactions, such as deployments and key rotation, still crosses to Akka. See Akka Automated Operations technical overview for the architecture and data flows, and Choosing your environment for sizing, availability and data residency.

What Akka Automated Operations does

Akka Automated Operations (AAO) runs the runtime for you in the Serverless, Dedicated, BYOC and BYOK8s options. It handles the work that comes after deployment:

  • Elastic scaling: capacity follows load.

  • Multi-region operations: one system across regions. When a region is lost, an operator fails over to another region. Replication is asynchronous, so writes that the lost region had not yet replicated are not available in the others.

  • Rolling updates: new versions roll out in waves, including changes to how services store data.

  • Live patching: runtime and security patches, and certificate and key rotation, without a redeployment.

  • Moving between environments: cloud to cloud, from Akka’s cloud into your own account, or back on premises, by replication into a new region.

  • Zero-trust networking: every workload is attested rather than trusted for its network location. In your own cloud account, services can also be operated as multi-tenant, with access controls across organizations.

  • Operations around the clock by Akka for the options Akka operates.

Self-managed operations give you the runtime, clustering, elasticity, resilience, and metrics, logs and traces. Logs, metrics and traces can be exported to OpenTelemetry, Prometheus or Splunk in every option. The other items above are AAO. See the feature comparison on Choose how you run Akka.

See concepts:deployment-model.adoc#automated-operations, Availability and recovery and Security and identity.

Where it is in the documentation

Section Start with

Getting started

Deploy to akka.io

Concepts

The runtime, Deployment and operations, Multi-region operations, Availability and recovery, concepts:deployment-model.adoc#automated-operations, Security and identity, Observability

Deployment

Choose how you run Akka, How AAO works, Self-managed operations, Production readiness

Operating

Operating, Services, Observe, Automate

Reference

Descriptors, Akka CLI

Part of the Akka Agentic AI Platform

  • Akka SDK components are what the runtime runs.

  • Akka Specify sets the operational envelope: service levels, regions and budgets.

  • Akka Verify enforces controls inside the runtime.

  • Akka Optimize serves models on accelerators in your regions.