Storage is the bottleneck, not compute
Radiology reads stall on retrieval. The array is healthy by its own metrics — the problem is throughput per study, not capacity or IOPS.
Solutions · Healthcare
Healthcare & life sciences
Imaging volumes grow faster than budgets, and patient data has strict handling rules.
What makes this sector awkward
Imaging and genomics produce data faster than storage budgets grow, the working set is large enough that caching does not help, and the handling rules around patient data constrain where that data can live and who can touch it.
High-throughput storage for imaging, a dual-fabric SAN that has been tested by pulling a path, self-hosted object storage with lifecycle policy, and controls aligned to the framework you are accountable under.
Four situations we see constantly
If two or three of these describe your environment, the audit is worth the money. If all four do, it is overdue.
The SAN has two fabrics that nobody has tested by pulling a path
Dual fabric in the diagram, single fabric in practice, because both switches sit in the same rack on the same feed.
Retention policy exists on paper
Nobody can say with confidence which studies are where, or prove that a retention rule is being enforced rather than hoped for.
Research and clinical data share infrastructure they should not
The separation was a VLAN and an agreement. It is not enforced at the platform layer, and it would not survive an audit.
Four phases, each useful on its own
Sequenced so that stopping after any phase leaves you better off rather than stranded mid-migration.
- 01
Measure the real workload
Baseline throughput per study and per read, not aggregate array IOPS. The aggregate number hides the tail that users actually experience.
You end up withA throughput baseline against imaging workflows, with the latency tail broken out.
- 02
Fix the fabric and the failure domains
Verify both fabrics are genuinely independent — power, uplinks, physical separation — then pull a path and time the failover.
You end up withA path-failover drill report with a measured switchover time and the remaining single points named.
- 03
Enforce separation at the platform layer
Move clinical and research separation out of policy documents into enforced boundaries: separate pools, separate credentials, separate networks.
You end up withA separation that holds when someone makes a mistake, not one that holds only when everyone remembers.
- 04
Make retention provable
Implement lifecycle policy in storage itself so retention is enforced by the system rather than by procedure.
You end up withAn artifact that answers "where is this study and why is it still here" without human archaeology.
The capability behind it
Each discipline has its own page with the technology stack, the acceptance criteria and an explicit scope boundary.
Distributed Storage Cluster Consulting & Operations
Ceph, BeeGFS and MinIO clusters designed around your real failure domains, benchmarked with fio, and proven to survive a disk, a node, and a rack going away.
Ceph (Squid / Reef) · Cephadm + Orchestrator · Rook-Ceph 1.17 · Ceph RGW (S3)Storage Area Network & Fabric Engineering — Fibre Channel, iSCSI and NVMe-oF
Dual-fabric SANs with zoning a human can read, multipathing proven by actually pulling a path, and NVMe-oF over lossless Ethernet that does not drop a frame under load.
Cisco MDS (NX-OS 9.x) · Brocade / Broadcom FOS 9.x · 32G / 64G FC · VSANsMulti-Cluster Fleet Management, Disaster Recovery & Security
The layer that holds every other cluster type together: fleets that deploy the same way everywhere, DR that has been rehearsed, and security that produces evidence.
Karmada · Rancher Fleet · Argo CD ApplicationSet · Cluster APIBackup, Restore and Disaster Recovery Cluster Services
Recovery tiers with a measured RTO, an off-site copy that cannot be encrypted by ransomware, and a restore drill the board can read.
Proxmox Backup Server (+ sync jobs) · Proxmox VE backup · Bacula / Bareos · AmandaHow this looks in practice
Media production · EU
A licensing renewal the board would not sign, turned into 140 VMs on open source
4 · waves, zero unplanned downtime
Read the engagement → Regional healthcare group · USFour nodes to eleven, with the storage rebalancing planned rather than survived
82% → 54% · cluster utilisation after expansion
Read the engagement → Public research university · EUA private cloud that tenants can self-serve, built to a failure-domain map
6 · tenants migrated off departmental islands
Read the engagement →Every case in this library is a representative engagement, written from our delivery patterns and marked as such. We do not present unnamed work as a named customer.
What we measure in this sector
These are the numbers we put in the delivery report. If we cannot measure it, we do not promise it.
- Storage throughput validated against real imaging workloads, tail included
- Path-failover drill with a measured switchover time
- Retention enforced by the storage layer, reportable as an artifact
- Enforced separation between clinical and research data
What we don't do
Every vendor says yes to everything. Here is where our answer is no, or where someone else needs to be in the room. If this list looks wrong for your situation, tell us — we would rather argue about scope before the contract than after.
- Out of scope
HIPAA compliance attestation and legal interpretation
We implement HIPAA-aligned technical controls for organisations holding their own BAA. The attestation is not ours to give.
- Out of scope
Medical device qualification and vendor certification
If a workload is a regulated device, its qualification belongs to the device vendor. We will build the platform it runs on, not certify it.
- Requires partner
PACS, RIS and clinical application administration
We handle the storage, fabric and platform underneath. Application-level configuration needs the application vendor, and we will say so before the contract.
- Included
Storage architecture, fabric engineering and lifecycle policy
Including the failover drills and the retention implementation.
Healthcare specifics
Our PACS vendor says only their reference architecture is supported. Can you work with that?
Yes. We will build to their constraints and tell you exactly where those constraints cost you throughput or money. Sometimes the honest answer is that following the reference architecture is correct; sometimes it is a compromise worth documenting.
Can you reduce cost without reducing performance?
Usually by changing the tiering rather than the array. Most imaging estates keep far more hot data than the clinical workflow needs, and moving the cold tail is where the money is.
Do you need access to patient data?
No. We need access to infrastructure, metrics and configuration. Where an investigation requires looking at actual content, we will say why, in writing, before doing it.
The engineering transfers, the constraints do not
Financial services & fintech Regulated, audited, and unforgiving of downtime. Every change needs evidence. Media & entertainment Render and transcode farms starved by storage, with peaks that double overnight. AI startups & ML teams Expensive accelerators, unstable multi-node training, and no proof of where the time goes. SaaS & internet platforms Multi-tenant isolation, unpredictable growth, and a cloud bill that scales faster than revenue. Research & higher education Slurm on one side, Kubernetes on the other, and a budget committee in the middle. Manufacturing & edge fleets Dozens of small sites, intermittent connectivity, and nobody on site who can debug a node.
Want this looked at against your healthcare environment?
Thirty minutes, no slides. Bring the awkward constraint — that is usually the interesting part.