The Cloud Isn’t a Silver Bullet for Healthcare; That’s a Problem
Remember when we all thought the cloud was going to save healthcare IT? A few years ago, executives were promising cost savings, increased agility, and security nirvana, all wrapped up in a neat, subscription-based package. Fast forward to today, and many healthcare CIOs are staring at the bill, scratching their heads, and asking one fundamental question:
Why isn’t this working the way all those airport billboards said it would? ??
Turns out, healthcare has a cloud problem, but not for the reasons you might think. It’s not about whether we should move to the cloud. That bridge already has water underneath it and we've crossed it about 2 miles ago. The issue is that we brought our baggage with us: decades of legacy investments, high customization needs, and security models that don’t translate well to cloud-native environments.
And now, we’re seeing the cracks.
A Sunk Cost Fallacy at Enterprise Scale
Hospitals are notoriously slow adopters of new technology, and for good reason. They can’t afford downtime. Lives are at stake. This means that every new investment is stacked on top of the old—a Frankenstein of infrastructure, patched together to ensure that no piece of the system breaks before its time. The cloud was supposed to fix that. Instead, it’s added complexity, not removed it.
I saw this firsthand when I worked on one of the earliest Epic migrations to an IaaS model (that I know of; I made no claim to factual accuracy on that one). The hospital system was seduced by the promise of cost savings and a fully managed infrastructure. The reality? A never-ending parade of hidden costs, budget overruns, and wildly inaccurate usage models.
And then there was the outsourced support model, where unplanned downtime "avoidance" took priority over quick resolutions. If something broke, you didn’t just reboot the system—oh no. That required approvals, ticket escalations, and hours of wasted time just to execute a simple fix that internal IT could have done in minutes.
Worst of all? Network performance crumbled under the weight of poor architectural assumptions.
You Can’t Route Every Patient Interaction Through the Cloud
Let’s talk about the biggest cloud networking mistake I’ve ever seen.
During that same migration, we were forced to shift all LAN-based communications onto a WAN, adopting what we'd now call a ?? zero-trust model ??, but at the time, it meant that every single service was tucked behind default-deny firewalls.
By design, some SQL calls take a LONG time. One day, we found a single SQL call took around 48 hours to complete, processing end of month processes.
Yes, you read that right. A database query that should’ve returned in milliseconds was now taking two full days to process. Why? Because long, stateless SQL calls don’t play nice with stateful firewalls. The state table tracking the connection kept timing out, effectively forcing the system to start over, again and again, leading to more server resources utilized, resulting in ONE of many VERY LONG repeated queries....
And let’s not forget vMotion—one of VMware’s crown jewels for VM mobility. Back then, it required Layer 2 adjacency, meaning VMs couldn’t be moved between hosts without some serious network gymnastics. Trying to extend Layer 2 across a WAN? Not fun.
Today, VMware has evolved vMotion, allowing for Layer 3 migrations, but it still requires careful planning. Other hypervisors like Microsoft Hyper-V, KVM, and Xen have their own versions of live migration, each with unique latency and bandwidth limitations.
How Do Other Virtualization Platforms Compare?
It really doesn't matter, unless you've locked into VMware ecosystems to solve architectural and performance problems....and if you're in Healthcare, you likely did if you're running Epic.
However, it speaks to the suitcases of legacy debt we can't simply shake, or even get until the weight restrictions to avoid those extra travel fees.
领英推荐
The NeverEnding (Cloud Migration) Story ??
Epic’s move to the cloud sounds great on paper, until you hit the database problem. See, in most industries, shifting to AWS means just pick a managed database service and go;Aurora, RDS, DynamoDB, Redshift. Player's choice. But Epic doesn’t work like that. It relies on InterSystems IRIS, a highly specialized, high-performance, multi-model database that doesn’t have a cloud-native equivalent.
?? Translation: If you want to move Epic to AWS, you don’t get to use AWS’s fully managed database services. Instead, you must manually run IRIS on EC2, reconfiguring networking, storage, and performance tuning, just like on-prem.
So instead of fully adopting AWS’s “managed-everything” model, you end up paying for cloud resources while still managing your database the old-fashioned way. It’s the worst of both worlds. And just when you think you’ve solved one migration problem, another pops up like a villain in a 1980s fantasy movie, refusing to die. Like The NeverEnding Story, the cloud migration journey never actually ends.
Or maybe it's more like The Song That Never Ends: ?? It just goes on and on, my friends… Some CIOs started moving, not knowing what it was, and they’ll keep on migrating forever just because… ??
(And if you just read that to the tune in your head, you’re welcome.)
Final Thought: Why the Cloud is Just a Stop, Not the Destination
If AWS had a managed database service for Epic, the migration conversation would be different. But they don’t. So healthcare IT teams have to take on significant operational overhead—which could offset the expected cost savings of cloud migration.
And in my recent quantitative risk analysis articles of HIPAA’s new security requirements, cost savings kept surfacing as a central theme. After all, risk management isn’t just about security; it’s about avoiding unnecessary losses. And in healthcare, that includes financial losses:
But here’s the thing: It doesn’t matter where you shift your spend if you end up spending more than the original solution. Healthcare wants to get out of the data center business, and honestly, I get it. But it's not really feasible on public infrastructure that has no viable backup.
Data centers in hospitals aren’t just a cost center; they’re critical infrastructure. Just like backup generators, they exist because failure isn’t an option. And it’s time we stop trying to cost-avoid our way around that reality. Until AWS and InterSystems build a native cloud solution, most hospitals will struggle to move beyond lift-and-shift migrations, and many may choose to stay hybrid for the foreseeable future.
?? Cloud for healthcare isn’t a destination—it’s just one stop on a much longer road trip.
One that, much like The NeverEnding Story, never really ends.
?? What’s been your experience moving EHRs to the cloud? Have you faced the IRIS database roadblock? Let’s discuss! ??
?? The NeverEnding Cloud Migration Story: You can move Epic to the cloud… but should you?
Amazon Web Services (AWS) VMware InterSystems Software Development at Epic Aaron Miri HIMSS HealthcareITNews BeckersHealthcare
?? Have you faced these challenges? What’s your take on EHRs in the cloud? Let’s discuss!