AWS Architecture Blog·September 17, 2026Building a Cloud-Native PACS on AWS for Scalable Medical Imaging
This article outlines a hybrid cloud architecture for Picture Archiving and Communication Systems (PACS) on AWS, designed to overcome the limitations of traditional on-premises solutions in multi-hospital networks. It details a hub-and-spoke model that centralizes DICOM image storage in the cloud while maintaining local PACS instances for immediate access, enhancing scalability, reducing costs, and improving data accessibility across facilities.
Read original on AWS Architecture BlogTraditional PACS architectures, relying on on-premises SAN/NAS at each hospital, face significant challenges in multi-hospital networks due to the ever-increasing volume of medical imaging data (50-200 TB/year per network, with 7-10 year retention). These challenges include escalating storage costs, data silos preventing cross-facility access, and bottlenecks in radiologist reporting due to lack of load balancing across facilities. A cloud-native or hybrid cloud approach is essential for modernizing these systems.
Core PACS Components and Their Cloud Adaptations
Regardless of deployment, a PACS system comprises six core components. In a cloud migration, these components are re-hosted and enhanced with cloud-native capabilities.
| Component | Role | Cloud Adaptation/Benefit |
|---|
Hybrid Cloud Architecture: Hub-and-Spoke Model
The proposed architecture utilizes a hub-and-spoke model. Each hospital operates a local PACS instance (spoke) for immediate image access, connected to a centralized cloud archive (hub). This setup ensures low-latency access for daily workflows while providing a unified, scalable, and resilient data store for the entire network. AWS Direct Connect or Site-to-Site VPN facilitate secure, high-throughput connectivity between on-premises spokes and the cloud hub.
- Local Spoke: Each hospital retains local Web, VNA, Application servers, and a local database. Modalities send DICOM objects to the local VNA, ensuring studies are immediately available at LAN speed.
- Asynchronous Cloud Replication: New images are asynchronously replicated from local storage to Amazon S3 in the centralized cloud archive, preventing workflow blockages.
- Centralized Cloud Hub: The cloud hub runs the same PACS vendor software across two Availability Zones, leveraging Amazon EC2 for Web, VNA, and App Servers behind Network Load Balancers. Amazon Aurora PostgreSQL serves as the highly available centralized metadata store with synchronous replication.
- Unified Patient Record: A common metadata database across all sites and the cloud enables seamless discovery and retrieval of studies from any facility, creating a unified patient imaging record.
- Transparent Viewer Experience: Clinicians interact with a single interface. If a study is locally cached, it's served at LAN speed. If not, the cloud viewer streams from Amazon S3 via Amazon CloudFront and AWS WAF, abstracting the data source from the user.
Key System Design Takeaways
This architecture demonstrates a robust strategy for migrating critical, high-volume on-premises systems to a hybrid cloud. It balances performance requirements (low-latency local access) with scalability, cost-efficiency, and disaster recovery benefits of the cloud. Asynchronous replication and a unified metadata layer are crucial for maintaining operational continuity and data consistency across the distributed environment.