Healthcare System Deploys Proxmox VE with Independent Storage Scaling

Organization

A US-based multi-campus healthcare organization operating an acute care hospital, outpatient surgical centers, and specialty clinics serving over 60,000 patients annually. With 850+ employees, the organization runs its electronic medical records (EMR) system, PACS image archive, clinical applications, and backup infrastructure on a shared virtualized environment.

Industry

Healthcare

Challenges

The organization had been running a legacy three-host hypervisor cluster that was approaching end-of-support. As the IT team evaluated replacement platforms, they recognized that the real constraint was not the hypervisor choice alone – it was the storage architecture underneath. The EMR database VMs demanded low-latency IOPS, the PACS archive needed sequential write throughput for radiology images, and the backup repository required hundreds of terabytes of retention capacity. The existing all-flash SAN served all three, but the cost of expanding it for the backup workload alone was prohibitive.

The team evaluated hyperconverged infrastructure as a replacement, but the scaling model was a poor fit. Adding HCI nodes to meet the backup repository’s capacity requirements meant purchasing compute and RAID processing that the EMR and PACS workloads did not need. The cost per usable terabyte for capacity expansion in an HCI model included CPUs, memory, and licensing that had no bearing on the backup workload’s actual requirement.

“The problem isn’t the hypervisor – it’s that storage, RAID processing, and compute are sold as a bundle. We need to grow our backup capacity by 120 TB, and HCI says we need to buy two new nodes with CPUs we’ll never use.” – Emily, Director of IT Infrastructure

The third concern was performance isolation during storage maintenance. When a drive in a traditional storage system failed, the rebuild consumed I/O bandwidth that degraded VM performance for the EMR and PACS workloads – both of which are latency-sensitive during clinical hours. The IT team needed an architecture where a drive rebuild or a scrub in the backup tier did not affect production database I/O.

Solution

The organization selected the StoneFly USS Proxmox VE Enterprise Appliance with Totally Disaggregated Architecture. The appliance ships as three integrated nodes: an HCI controller running Proxmox VE with active-active RAID access, a dedicated hardware RAID controller appliance, and an EBOD providing raw capacity. Each node handles one function – compute, storage processing, or capacity – and each scales independently.

The HCI controller presents LUNs through active-active RAID paths, with both controllers processing I/O simultaneously, doubling the aggregate throughput available to VM disk operations. The dedicated RAID controller offloads every parity calculation, scrub, and rebuild from the HCI controller, so VM performance remains consistent during storage maintenance events. The EBOD is configured with separate drive pools – NVMe for the EMR database LUNs, SAS for PACS image storage, and high-capacity SATA for the backup repository – all managed through a single RAID controller and presented to Proxmox VE as independent LUN targets.

“The TD architecture was the deciding factor. We can scale backup capacity by adding an EBOD shelf – no new HCI controller, no new RAID controller, no compute cost we do not need.” – Emily, Director of IT Infrastructure

Results

The deployment gave the healthcare organization a Proxmox VE environment where compute, storage processing, and capacity each scaled on their own terms – matching the divergent growth profiles of its EMR, PACS, and backup workloads without forcing a compromise on any of them.

Independent Scaling Eliminated Compute Overprovisioning

The backup repository reached 80 percent capacity within six months of go-live. Under the previous architecture or an HCI alternative, expanding would have required purchasing compute and RAID processing the organization did not need. With the totally disaggregated architecture, the IT team ordered and deployed a single 60-bay EBOD shelf – drives and a chassis only. The HCI controller and RAID controller continued operating unchanged, and the new capacity was presented to the Proxmox cluster within the same maintenance window.

Consistent VM I/O Performance During Storage Maintenance

Six drive rebuild events occurred across the storage tiers in the first year of operation. In each case, the dedicated RAID controller handled the rebuild entirely within its own RAID processor and onboard cache — reading surviving drives, calculating parity, and writing to the replacement drive. The HCI controller’s processor cycles remained dedicated to running VMs. The EMR database, which would have experienced measurable I/O latency degradation under the previous architecture, showed no performance variance during any of the rebuild events.

“The first time we had a drive fail during clinical hours, I braced for the support calls. They never came. The rebuild happened on the RAID controller – our VMs did not even notice.” – Emily, Director of IT Infrastructure

Tiered Storage Matched Each Workload’s Requirements

The NVMe pool serving the EMR database delivered sub-millisecond latency for transactional workloads. The SAS pool provided consistent sequential throughput for PACS ingestion and retrieval. The SATA pool in the high-density EBOD provided cost-effective capacity for the backup repository at the lowest cost per terabyte in the environment. Each workload consumed storage from the tier designed for its access pattern, and consolidation onto a single appliance eliminated the management overhead of three separate storage systems.

HIPAA-Aligned Data Protection Without Performance Trade-offs

The active-active RAID controllers provided hardware-level redundancy for all three storage tiers, protecting patient data at the block level while maintaining aggregate throughput from both controllers simultaneously. Proxmox Backup Server repositories were configured on the SATA tier, meeting HIPAA’s data integrity and retention requirements for backup data without competing with production I/O for controller or cache resources. The dedicated RAID controller’s battery-backed write cache ensured ZFS consistency for the EMR database volumes, with no risk of write acknowledgment before data reached stable storage.

Single-Window Management Across Compute, RAID, and Capacity

The IT team managed the entire environment – Proxmox VE on the HCI controller, RAID configuration on the dedicated controller, and EBOD inventory – through the StoneFly unified web-based management console. VM provisioning, storage pool expansion, and RAID monitoring did not require separate tools or vendor portals. When the organization’s annual compliance audit reviewed the backup infrastructure, the consolidated management and reporting interface generated the required storage and data protection documentation in under an hour.

Looking to deploy Proxmox VE with storage that scales independently of compute? Contact Us Today!

Contact us to discuss your Proxmox VE project requirements at [email protected] or call +1 510 265 1616.