I love cloud, so any step towards it I personally see as being truly great. But all choices have consequences and with this article, I want to highlight those of moving to private cloud.
A private cloud solution vendor can easily show you a calculation showing how much more cost efficient it is compared to public cloud, and of course, there are scenarios where that is true. If you have a set of applications that only require specific infrastructure services, the life cycle of those apps is known and predictable and your growth is also known and predictable. Ultimately, an Onprem Hyper Conversed solution with some private cloud features is probably cheaper, at least for a period of time.
However, the private cloud costs start to stack up over time…
Who does the R&D to maintain your private cloud stack (the software functionality to automate your infrastructure)? Take OpenStack, VMware stack, or any other way to build the needed cloud functionality. You can have over 20 software components each that have their own life cycle, and any update needs to be tested against all other components. Vendors are addressing this by packaging the components and doing the testing for you, but then you are tied to their cycles and version choices.
What does it cost to build additional services beyond basic IaaS features? Continuous delivery pipeline features which integrate your favourite tool to your private cloud, automated provisioning of entire servers including database and app servers, PaaS features, machine learning features, and these are just the bare minimum.
There are of course, the costs of planning capacity and maintaining needed as a buffer for growth. But a far bigger cost is the tech refresh cycle with much needed migrations and forced upgrades. Let’s assume you have a 4 year refresh cycle. Year 3 and 4 capacity upgrades are ridiculously expensive. What if the private cloud had room for year 5 growth? How much does the maintenance extension cost for year 5? The list goes on, but I still have never seen calculations considering these unpleasant surprises which are, in most cases, inevitable (who knows how much capacity we would need in 4 years!)
Multiple cloud costs
There is no single cloud stack (hardware and software) that covers all your needs. The requirements of basic VM IaaS to SAP S4 Hana to Machine Learning are very different on all levels. As a basic example, ML requires GPUs, SAP requires a huge amount of memory etc. There is no practical way to deliver that from one cloud, so be prepared to build and manage multiple private clouds.
Consider all the costs of maintaining and developing your private cloud. Then consider the situation 4 years from now and the money you could potentially spend on that. Wouldn’t you have been better off spending that money modernising your applications and creating new digital innovation for your business and customers?
The cost of these in public cloud? You only pay for consumption, not for development, maintenance etc. The list goes on…