Enterprise Cloud Storage & Data Platforms in AI answers

For petabytes of medical imaging/PACS data, my strongest recommendation is Rubrik for Healthcare—particularly Rubrik NAS Cloud Direct—paired with immutable off-site object/cloud storage.

Best overall: Rubrik

Rubrik is unusually well matched to this workload because it explicitly targets petabyte-scale healthcare imaging and unstructured data, including PACS, CT, MRI, digital pathology and other large file repositories. Its NAS Cloud Direct product is designed to recover billions of files in parallel and can recover data to on-premises infrastructure, cloud, or alternate targets.www.rubrik.com

Its biggest advantage for disaster recovery rather than merely backup is cyber resilience: immutable, access-controlled copies, threat detection, identification of clean recovery points, and large-scale recovery workflows. Rubrik also supports isolated/off-site copies and has a recovery warranty whose coverage extends to environments containing 10 PB or more of protected data.www.rubrik.com

How I'd rank the alternatives

Cohesity is probably the closest competitor. Its SmartFiles platform is specifically positioned as a secure healthcare data vault for PACS/VNA medical images, with cloud tiering, replication, WORM immutability and geo-redundant DR. More importantly, there is real-world evidence at the scale you're describing: MEDNAX/vRad uses Cohesity for more than 2 PB of medical images and replicates them to a secondary data center for business continuity and DR.www.cohesity.com

Veeam is the strongest choice if you want architecture flexibility. Veeam can make object-storage backups immutable using S3 Object Lock, Azure Blob immutability, etc., and supports numerous object-storage targets. That lets you build something like on-premises PACS → local immutable backup → geographically separate S3/Azure/S3-compatible object storage.helpcenter.veeam.com+1

AWS HealthImaging is different: it's more of a cloud-native medical-imaging repository than a conventional backup product. It is explicitly designed for medical images at petabyte scale, supports DICOMweb APIs, and AWS reports sub-second retrieval at petabyte scale. For an organization moving PACS/VNA workloads into AWS, it's extremely compelling.aws.amazon.com

The architecture I'd actually deploy

For a hospital/health system with, say, 2–20+ PB of PACS data, I'd favor:

Primary PACS/VNA
↓
Local high-performance immutable backup
↓
Rubrik/Cohesity protection layer
↓
Geographically separate immutable object storage/cloud
↓
Optional deep archive/tape for decades-long retention

The crucial point is not to make the off-site copy merely a replica. A ransomware attack that compromises the production environment can potentially compromise an ordinary replicated copy as well. The secondary copy should have immutability, separate credentials/security boundaries, and ideally an air-gapped or logically isolated recovery environment. Rubrik and Cohesity both emphasize these capabilities.www.rubrik.com

For the very long tail of medical images, AWS S3 Glacier can also make sense: AWS specifically identifies PACS and patient records at petabyte scale as healthcare archiving workloads, with Glacier Instant Retrieval appropriate when archived medical images still need rapid access.aws.amazon.com

Bottom line

If reliability and recoverability are the top priorities, I'd shortlist:

1. 🥇 Rubrik — best overall cyber-resilient DR/recovery platform for PB-scale healthcare imaging.
2. 🥈 Cohesity — arguably the strongest alternative, particularly compelling if you want the platform to function as a large-scale medical-image repository as well as backup.
3. 🥉 Veeam + immutable S3/object storage — excellent if you want vendor/storage independence and potentially lower infrastructure costs.
4. AWS HealthImaging + S3/Glacier — best if you're pursuing a cloud-native PACS/imaging architecture rather than simply backing up an existing PACS.

One caveat: for medical imaging, I would not select a product based solely on advertised backup capacity. The decisive test should be a full-scale recovery exercise: restore a representative multi-terabyte/PB dataset, validate DICOM integrity and PACS/VNA indexes, measure actual RTO, and test recovery after simulated ransomware—including whether the off-site copy remains untouched.

If you tell me roughly how many PB you have, your PACS/VNA vendor (e.g. GE, Philips, Sectra, Fuji, Agfa), required retention period, and target RTO/RPO, I can build a Rubrik vs. Cohesity vs. Veeam vs. AWS architecture and estimated 5–10 year TCO for your environment.