If you haven’t yet, I urge you to read “The Big Cloud Exit FAQ” by David Heinemeier Hansson of 37signals. What struck me was his candor about why the company moved off the cloud and his responses to the criticism and questions about this decision.
The simple fact is that public cloud platforms are not suitable for some workloads or data sets, which can live on more cost-effective platforms such as your own servers in your own data center. This is not a knock on cloud computing; it’s just the reality. The cloud can be costly, and it’s not economically viable for some applications and data types. Also, hardware, including storage and computing, has gotten a lot cheaper, as Hansson points out, as did I in my latest book.
Facts don’t care about your bias
What was most interesting about Hansson’s article is his comment, “To say this journey was controversial is putting it mildly.” The decision to put their applications and data back on traditional hardware caused a great deal of second-guessing and fighting the cloud bias out there.
I’ve been taken to task anytime I suggest non-cloud platforms, whether I’m speaking at conferences, sitting on panels, or even in meetings when I thought I was in the company of other professionals. The bias against cloud computing I ran into years ago has shifted to “give me cloud or give me death” without considering any requirements.
I would not be doing my job as an architect if I were not open to anything that works best for the business, regardless of some people’s silly bias. Many don’t see past the last big cloud conference’s cocktail parties to understand that just tossing stuff into a public cloud won’t always lead you to the most optimized use of cash.
Indeed, when the applications and data sets are simple or they support a narrow set of services, such as providing a single tool on demand, the cloud is usually not the right choice. Even if it works, it will cost way too much and not provide any additional advantages over running the workloads and data sets on premises.
More complex deployments that mix hundreds of different types of services, such as AI, data, security, etc., are often worth the investment in cloud. I’ll say it again: The cloud is not the right solution for many use cases, and our ability to figure that out now will pay huge dividends.
I expect we’ll see more case studies, such as Hansson’s. We’ve already seen dozens of technology providers that were “born in the cloud” change to traditional on-premises infrastructure. Some announced it, but many did not. In all of those cases, it was not that the cloud did not provide the quality of services they needed—it did. It was just way more cost-effective to consider the alternatives.
What 37signals found
37signals was a significant cloud user with a $3.2 million cloud budget for 2022. The company pledged $600,000 to procure Dell servers, envisioning significant savings during the next five years.
Of course, there were questions, and Hansson did an excellent job of addressing them one by one in the FAQ, such as the additional costs in terms of humans needed to run the on-premises systems, how optimization only took them so far in the cloud, and how they handled security requirements.
Hansson also explained the limited abilities of cloud-native applications to reduce costs and highlighted the need for a world-class team to address security concerns, which the company has. Notably, privacy regulations and General Data Protection Regulation compliance were underscored as reasons for European companies to opt for self-owned hardware as opposed to relying on the cloud. Of course, this is not the case for everyone.
Before the move, 37signals made significant efforts to address reliability, performance, and cost comparisons. Through financial comparison and a wise investment in hardware, the company realized immediate paybacks and positioned itself favorably for future technological achievements. In other words, it’s just as good, it’s cheaper, and they have more direct control, all of which returns value to the business.
Cloud is not always the answer
Everyone is looking for a single answer, and it doesn’t exist. The requirements of your systems will dictate what platform you should use—not whatever seems trendy. Sometimes the cloud provides the most value, but not always.
There is a lot of mythology out there that the cloud is always less expensive, more reliable, and more secure. Sure, you can make that case, and I have made that case. Cloud is where most innovation is occurring now, and those who remain on premises are going to see that they don’t get as much love as cloud services do. That alone may be reason to move to or stay with the cloud.