By Charles E. Christian, CHCIO, CDH-E, LFCHIME, LFHIMSS, Vice President of Technology & CTO, Franciscan Health
I’m working on my 6th decade in healthcare, so I’ve been around since before data centers were a thing in this industry. I watched the transition from mainframes to minicomputers to server clusters and now to virtualized servers; it looks like circular evolution of how computing resources are provided. Therefore, I must ask, are we looking at continuing the evolution when datacenters are again not really a thing in healthcare in a few years?
With the increased functionality of cloud-based computing/storage resources, and increased network bandwidth availability (with associated lowering of network latency), applications are rewritten to take advantage of hyper-converged environments, and the potential for decreased costs; all these things make moving workloads to the public cloud infrastructure more appealing or at least that is the current thinking.
That said, moving workloads to a consumption-based cloud model may have some unexpected consequences. Throughout my career, I’ve worked to provide application specific costing/expense models, which can prove difficult at best in the standard datacenter setting; estimates and assumptions are used and compounded by the ability to utilize hardware well past their projected capitalized/depreciated life. I’ve learned that when you are purchasing capital equipment that will be housed in your datacenter, that purchase only appears as part of the aggregated capital asset and depreciation schedules. However, when you move to a cloud-based consumption model, it becomes an expense that appears on the monthly and annual expense reports; basically, putting a bright spotlight on the cost of cloud services and a target to be controlled/decreased. It’s a gift that just keeps giving; you get the pleasure of explaining that line item monthly and annually.
Building out or retrofitting an existing datacenter can be time consuming and require additional and unexpected costs related to building code changes and required upgrades.
The other part of moving workloads to the cloud is related to being a good steward of your organization’s funding; if you have the capacity within an existing ESX host sitting in your datacenter and you can spin up the needed servers where there is almost no ongoing cost (other than OS, etc.), why would you move it to the cloud and begin paying monthly consumption costs. This is the mental challenge and internal debate that I continue to have.
I have not always been a fan of the cloud environment option; color me a little paranoid regarding HIPAA and other privacy regulations and the fact that we would be storing PHI (personal health information) in that cloud. In the early days of the public cloud, the healthcare providers were still on the hook in the event of a data breach; those rules changed several years ago, which moves some of the risks onto the cloud provider; however, contracting still attempts to move the risk to the organization whose data is being hosted and not so much with the hosting company. The devil is always in the details of any contract and/or agreement.
In the last several years, I can’t think of a single CIO and/or CTO that is not actively looking to move workloads to the public cloud. Many of these same people used to have a “cloud centric” or a “cloud first” approach, that is, until they had the opportunity to fully experience the costs of just moving anything and everything to the cloud. Please note that I do think that the public cloud should be a part of any workload placement strategy; however, if you don’t have the correct governance and cost monitoring processes in place, the experience might be painful. The cost models for cloud-based services are complex for a reason, very much like software licensing; as soon as you think you have the model nailed own, the rules change, or so it seems.
I’d also like to note that in the beginning, it should be your goal/task to establish what “cloud” means to your organization. To some, having resources in the cloud just means “not in my datacenter.” That is one way of looking at it, but I’m certain that we’ll all agree that it’s just not that simple. You have SaaS, IaaS, PaaS, etc. options, all of which fall into the “not in my datacenter” definition but have very different cost options/models and different levels of associated service offerings. This is where the workload placement strategy comes into play, placing the workloads where they are best suited from platform and cost perspectives.
So, let’s swing back around to the datacenter itself; there are some folks that believe you do not need a datacenter today. To some extent, I will agree; however, in good Six Sigma fashion, I also think it depends. Many of the applications that we run in healthcare today are very dependent upon low network latency, therefore, they must be either on-prim or in close proximity or setup in an edge computing type of environment. These legacy applications are getting fewer each year, but they still exist, and some may be critical components of your clinical solutions.
For me, it all boils down to finding the balance between functionality, requirements, and costs; there’s always a three-legged stool in almost any story. I believe looking at what needs to be accomplished is always a great starting point; if you need to make a large capital investment to get your datacenter ready for a new solution, then standing that up in the public cloud may be a better use of time and funding. Building out or retrofitting an existing datacenter can be time consuming and require additional and unexpected costs related to building code changes and required upgrades. Building new subscriptions in an existing public cloud environment can be accomplished in much less time than purchasing and installing new hardware in your own datacenter.
My last and parting thought is around the hardware manufacturers and suppliers; with more and more infrastructure being supplied by the various public cloud hosting companies, will this have an impact on the companies like HP, Dell and others who typically sell to a healthcare organization? The short answer is, probably; unless they are now supplying hardware to the cloud hosting companies. I’m guessing that we will need to see how that plays out and if these companies fall to the same faith as Data General, NEC, etc. See, I told you, I’ve been around a long time.