“Hamilton’s Law”
By Scott Hamilton
It has been a while since I talked about this topic, but I had a prediction, let’s just call it “Hamilton’s Law” just for fun, that states on an approximately 10-year cycle companies make a shift from centralized computing to decentralized computing and back. The last shift was the start of cloud services becoming mainstream; this was between 2010-2015 depending on which source you use for the date. This was the first shift back from decentralized computing in about a decade. So I guess for you to understand how “Hamilton’s Law” works, you must first understand the definition of centralized versus non-centralized computing.
Centralized computing means that a bulk of the workload done in the general office setting is done on computing resources outside of that office. In the early days of computing, work was done on a terminal in the office which linked to a mainframe where all the actual computing power lived. The cloud was a shift to centralized computing because, since its inception, more and more of our data and utilities have been web-based. We even type our documents on Google Docs, or Microsoft Office 365, in the cloud today. So it is safe to say we are operating in a centralized computing environment. We can take it a step farther as we see many major corporations depending on cloud services like Google, Microsoft, and Amazon for maintaining their network infrastructure; the corporate data center seemed to be a thing of the past. Especially over the course of the last five years.
Decentralized computing is where a bulk of the computing workload is local to the business, or even local to the desktop. Prior to the cloud migrations in 2010-2015, most of our daily work was done on our own computer, be it a desktop or a laptop; our documents were either stored locally, or at most, sitting on a file server in our office or one of our corporate data-centers. This was a decentralized system in that our tools and data were scattered across various systems within the company and nothing really left the control of our own information technology departments.
There were two main reasons behind the rapid migration to the cloud over the last several years. One was the convenience of being able to access our data from anywhere in the world securely and reliably. Companies gained the benefit of cloud service providers’ large global networks and data centers, which allowed them to replicate business-critical data not only across the state, but around the world. This provided a level of data redundancy that was never possible before without spending millions on replicating infrastructure. The cost for the cloud services were way less than the benefits provided. While running services in the cloud and storing data in the cloud was more expensive than doing it locally, the real savings was in comparing the cost to the benefit.
These reasons for a shift, ironically, were the same reasons for every historic swing between centralized and decentralized computing. As the cost of computers shifted, companies could afford more computing power (decentralized) or less computing power (centralized). The cloud changed things a little bit here as companies just paid for the amount of computer storage that they used, so the cost of computing systems were less of a factor in this shift and it was more about the benefits of this technology. Cloud technologies, as far as the software behind them, have become so mainstream that some of the bigger companies started to build their own “clouds,” partly out of worry about data-security, and partly out of the rise in costs for cloud services.
We have seen several companies drop multi-million dollar cloud contracts in the last six months, as they have migrated their business critical data and operations back into company owned data centers. This is the approximately 10-year cycle of decentralization of “Hamilton’s Law.” Only this time it was not driven by cost of computer hardware, but by risk of business. There are two main factors driving this shift and I will cover each of them briefly here, and might write some future articles giving more details. The latest statistics show 70-80 percent of businesses are making the shift away from the cloud, at least for some of their existing cloud workloads.
The first and probably main reason for this shift is a rather hidden expense of operational overhead. Early on, the cloud was providing most of the hard work when it came to upgrading software, maintaining systems and keeping things generally operational. Over the past few years, major cloud providers have begun to shift some, if not all of this operational complexity back on the customer through disabling some historic cloud features, placing higher price “digital tax” on other features, and putting companies in a perpetual migration cycle of moving workloads from one cloud service to another. This has vastly increased the operational overhead costs to companies using the cloud. A really big one is the cost of data transfer to and from the cloud.
This brings up the second point; early on cloud service providers only charged for the computers and data storage and did not charge for network connectivity and data transfers. This changed about five years ago, first with Google cloud services starting to charge fees for data egress. Egress is the process of moving data from the cloud to a local data-center or other cloud provider. They still allowed ingress, moving data in, for free. Two years ago that changed and they now charge for data transfer outside their infrastructure in both directions. We see the cost effectiveness of the cloud shifting as navigating the customer agreements has become cost prohibitive; as a result companies are making a shift back to decentralized computing nearly exactly as predicted by “Hamilton’s Law.”
Until next week, stay safe and learn something new.
Scott Hamilton is an Expert in Emerging Technologies at ATOS and can be reached with questions and comments via email to shamilton@techshepherd.org or through his website at https://www.techshepherd.org.
