The healthcare industry is unique in so many ways.
To begin with, it lacks a common objective function on which to optimize. This is because different healthcare organizations, both payers and providers, value different things. Profitability is not the universal goal - there are teaching hospitals, faith-based hospitals and not-for-profit insurance plans. Consider also that a large portion of the industry is guided by a 2,500-year oath at the point of interaction. Further, healthcare is at once highly regulated and process-driven yet intensely personalized. The list goes on with each example underscoring how different healthcare is from other industries.
Despite the uniqueness of the industry, it does share a common challenge with every other industry - that of managing and extracting value from their data.
Healthcare produces a disproportionate share of the world’s data - estimated at 30% of electronic storage. More importantly, the rate of growth is stunning. An IDC/DellEMC report found that between 2016 and 2018, data grew from 1.45 PB to 9.70 PB or 878%. Assuming that growth rate has not abated - which is conservative - those healthcare organizations would now be managing close to 100 PB each.
The volume of data and the attendant growth rate have required the healthcare industry to radically re-evaluate their storage architecture. What was a solved problem four years ago is now the primary challenge facing the industry - even larger than analytics.
The most sophisticated payers, providers and technology vendors have come to the realization that object storage, whether in the cloud or on-premises is the answer.
Why Object Storage is now the Foundational Storage Tier in Healthcare
The primary reason for the shift from SAN/NAS to object has to do with the aforementioned scale. Those legacy systems were not designed to perform at this scale. Further, no database can fit 100 PB, even if the data were arranged neatly as columns and rows, which it is not. Modern healthcare data is unstructured - images, scans, log data, event streams.
While the unstructured nature of data and the volume is a key driver for object storage, it is not traditional object storage that healthcare is adopting. Traditional object storage was not designed for speed or to be cloud-native or application-oriented. It was designed for archival or backup.
Amazon S3 changed that. They proved that object storage could be fast and cloud native. Furthermore, with the advent of S3 Select, AWS proved that object storage could also be searchable. This was a game changer.
Object storage could now serve as the foundational component of healthcare’s storage architecture. Consider the following. Modern object storage is inserted in the list below to distinguish between traditional, archival appliance-based object storage and what is sweeping the healthcare industry:
- Object storage scales without complexity.
- Object storage is designed for unstructured data.
- Object storage offers vastly superior economics vs. block or file.
- Modern object storage API is RESTful (based on Amazon’s S3 API) making it compatible with cloud native applications.
- Modern object storage is searchable using standard SQL.
- Modern object storage is fast. In the case of MinIO, it is hyper-fast, delivering speeds in excess of 171 GB/s on 32 nodes of NVMe and 9.4 GB/s on 24 nodes of HDD.
- Modern object storage is cloud-native and suited for containerization and orchestration, making it truly elastic.
For these reasons and others, modern object storage has become the default storage architecture in healthcare. It will take time to migrate away from legacy applications that depend on SAN and NAS, but not as long as one might think. Those solutions cannot scale. They have to be replaced - it is a business imperative, not a technical decision.
The New Edge Model + A 360 Patient View
Healthcare data is created at the edge and shared across the organization. This is simply how it works. MRIs, CTs, X Rays, ultrasounds are taken at the edge, are digital from inception, are shared in real time with the doctor and ultimately stored somewhere more central. Almost every single interaction from labs to IoT data from kidney dialysis machines follows the same model: data is collected or generated at the edge, made readily available to the organization and ultimately stored elsewhere.
Modern object storage is inherently built to share data efficiently across the Internet. SAN/NAS systems are designed for low latency applications within the data center. Consider the fact that even on AWS, you cannot access EBS (Elastic Block) or EFS (Elastic File) from outside the cloud. It is just not set up for that.
This modernization of the architecture has massive implications for the cost of healthcare. By moving the legacy IT infrastructure out of the individual hospitals into a common shared cloud infrastructure, all of the benefits of containerization and orchestration come into play.
By developing dense object storage clusters at the edge, the information can be processed and shared locally and eventually migrated to a lower tier, either in a data center or into the public cloud if the security and resilience meet requirements.
The Application Economy + Decline of File and Block
Healthcare is an industry defined by applications. Epic is an application. Cerner is an application. IBM Watson is an application. Every insurance company has dozens, if not hundreds, of applications - from population health to analytics. Every provider has dozens, if not hundreds, of applications from the EHR to chatbots.
Those applications require data. Modern applications communicate through RESTful APIs like Amazon’s S3 API. While many of those older applications may specifically target SAN/NAS environments with POSIX, they are rapidly giving way to RESTful solutions like S3.
Specifically, applications that run on the cloud (public or private) and at the edge are, by default, using RESTful APIs. Further, they can run in the cloud but access data at the edge - a critical model in healthcare. One only needs to consider the model that dominates today, an analytics application running on the cloud but accessing data that exists in the hospital. It is simply how modern software is architected.
Providers are not building data centers at each hospital location. They are using applications (speaking S3) to engage with data that is generated at the edge. Further, as we see more convergence between payer and provider (where each adapts skills from the other), we will see more of this application-oriented architecture and therefore more object storage in both places.
One last proof point: The foundation for interoperability in healthcare, FHIR, is based on RESTful APIs.
Security is another reason that modern object storage has flourished in healthcare. Modern object storage like MinIO or AWS has advanced encryption algorithms for data in flight or at rest.
In the case of a hospital, data loss is preferred over a data breach.
This, however, is not a choice a software architect should have to, or wants to, make. They want security and resilience.
Private cloud object storage can offer both. Using inline erasure coding and bitrot protection, MinIO is more resilient and more economical. More importantly, the fact that the data can remain within the control plane of the payer or provider insures against the types of breaches that have plagued the public cloud. Add to this key features like sophisticated encryption, tamper-proofing, object locking and legal lock for governance and compliance and you have a fundamentally more secure solution.
Public cloud vs. Private Cloud
We often get the question from our healthcare clients: public or private cloud? As a private cloud object storage provider our answer often surprises.
We tell our clients to start in the public cloud. There are skills to be mastered, best practices to be applied, lessons to be learned. Teams can develop expertise and adopt the mindset of the cloud. In the cloud they will learn about elasticity, containerization, Kubernetes, micro-services, best of breed for every component.
When those skills have been mastered - repatriate.
This ensures a private cloud that is based on the best practices of the industry while delivering superior resilience, security and economics.
Ultimately, data will exist everywhere. In multiple public clouds. On premises and in the public cloud. This requires a single API to speak across those disparate instances. As we have written above, that API is S3. MinIO’s credentials in that space are unquestioned. It is why Google and Microsoft both asked us to write S3 to Azure/GCP gateways for their clouds.
Experiencing Modern Object Storage
While our arguments will certainly resonate with healthcare technologists, most technologists want to experience it for themselves. With MinIO, that is a simple as downloading the code. MinIO is 100% open source, so the code you download is the same as our clients use. More than 200 enterprises across the healthcare spectrum use MinIO today, to power their products or for high performance object storage use cases. That list includes IQVIA, Unitedhealthcare, Cerner Corporation, McKesson, Stanford Hospital and Clinics, Kaiser Permanente, The Methodist Hospital, Epic Systems Corporation, National Institutes of Health,Fairview Health Services, BCBS of Oregon, North Carolina, Iowa, Illinois, Anthem, University Health Network and the Cleveland Clinic.
If you have questions, we have our legendary public Slack channel and our outstanding documentation to help. If you want to go deeper, reach out to us at firstname.lastname@example.org to learn more about how we can support your specific storage requirements.