Distributed Accelerator for Cloud Storage (DACS)

Distributed Accelerator for Cloud Storage (DACS) is a distributed caching system for data-intensive workloads on Kubernetes. It places a high-throughput cache between applications and remote object storage. Cache servers use node-local storage to reduce repeated remote reads and improve data access throughput.

DACS is intended for Kubernetes users who operate data-heavy workloads and for contributors working on the cache data plane, clients, or Kubernetes management components.

Important

DACS is under active development and should currently be treated as pre-release software. APIs, configuration, and deployment behavior may change.

Warning

Security model: DACS currently assumes a single trust domain equal to the entire Kubernetes cluster. It does not yet provide tenant isolation and is not suitable for multi-tenant clusters. Any workload that can reach a DACS endpoint is treated as fully trusted with all cached data and operations. Read the security model before deploying.

Choose your path

Get running in 10 minutes

Follow the AKS quickstart to install one ephemeral CacheServer on an existing AKS cluster and complete an end-to-end upload and download smoke test.

The published OCI charts and images at dacstest.azurecr.io are publicly readable and can be pulled anonymously.

Deploy DACS for production

Use the deployment user guide to prepare local NVMe/RAID storage, install matching chart and image versions, configure a cache, verify the deployment, scale it, troubleshoot it, and uninstall it.

Build and test changes

Use Build from source for a clean-clone build, targeted validation, build outputs, and common build failures. See Testing for integration, end-to-end, scenario, and performance test workflows.

How DACS works

Applications access data through CacheClient, the DACS Python client, or StorageIntercept. Cache servers store cache entries on local storage. On a cache miss, the storage interception path can read supported remote storage and populate the cache.

The Kubernetes cache controller creates and reconciles cache workloads from a Cache custom resource. The CacheServer scheduler places host-storage cache pods on suitable nodes. These management components are not in the data transfer path.

Application
    |
    +-- CacheClient / DACS client / StorageIntercept
          |-- cache lookup --------> CacheServer StatefulSet
          |                              |
          |                              +-- node-local storage
          |-- on cache miss: read --> Azure Blob / remote object storage
          +-- may populate cache ---> CacheServer StatefulSet

Kubernetes API
    |
    +-- Cache controller
    |
    +-- CacheServer scheduler

See Architecture for the detailed request flow.

Documentation map

Goal or component	Documentation
Find the project maintainers	Maintainers
Evaluate DACS on AKS	Get running in 10 minutes
Understand the system architecture	Architecture
Understand the security model and trust boundary	Security model
Install and operate DACS	Deployment user guide
Build from a clean clone	Build from source
Understand Azure Pipelines build caching	Azure Pipelines build and caching
Understand GitHub Actions build caching	GitHub Actions build and caching
Run tests	Testing
Understand artifact versions and releases	Versioning and releases
Use deployment tooling	Deployment workflows
Configure the CacheServer chart	CacheServer Helm chart
Configure the controller and Cache resource	Cache controller Helm chart
Understand controller scaling	Cache controller scaling
Use the Python client	DACS Python client
Configure storage interception	StorageIntercept

Platforms and source distribution

The build and runtime code targets Linux on amd64 and arm64. Documented developer environments include Linux containers and Ubuntu under WSL2. Kubernetes workflows cover Minikube and Azure Kubernetes Service (AKS). Native Windows and macOS builds are not currently documented or supported.

Published OCI charts and container images at dacstest.azurecr.io are anonymously pullable. Language packages are not currently published to public package indexes; build them from this source distribution.

Contributing, security, and support

• Read CONTRIBUTING.md before proposing a change.
• Follow the Microsoft Open Source Code of Conduct.
• Report vulnerabilities using the private process in SECURITY.md, not through a public issue.
• Understand the assumed trust boundary before deploying: see the security model.
• Use the channels in SUPPORT.md for questions, bugs, and feature requests.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos is subject to those third parties' policies.