There’s no shortage of news about the changing attitudes of enterprises toward cloud computing. Five years ago, it was all about “cloud-first” planning of applications, but recently “repatriation” of applications from cloud to data center has dominated, but both extremes are still represented. What’s the truth? One possible source that came to light this month is the evidence Amazon provided to the Competition and Markets Authority (CMA) in the UK, broadly equivalent to the Federal Trade Commission (FTC) here in the US. We still have to examine this material for bias (nobody lies to regulators, right?), and compare it to what I hear from enterprises.
The lead point of the submission was that Amazon “faces competition from on-premises IT and that analyst reports suggest that cloud services only make up around 15% of the IT services market as a whole” and that “the perception that once customers move to the cloud, they never return to on-premises is not correct.” Amazon says that 29% of all UK companies that left a public cloud provider went back to the data center, as opposed to another cloud provider, and that companies prefer the data center to cloud hosting, with only 12% being “cloud-first”. The submission, of course, was in response to a threat to apply regulations to cloud services in general, and to Amazon in particular as the leading cloud provider, but a totally fabricated CMA submission would surely pose a risk for any provider.
The submission says “AWS said that about 10 to 20 years ago, customers needed to make a massive capital investment spanning multiple months and years to migrate or switch on-premises data centre providers whereas today there is fierce competition in the cloud market and customers have more choice than ever before.” This sure sounds like the CMA is interpreting the cloud as an either/or with the data center, meaning that customers used to have build those massive data centers, but now the competitive cloud market has given them another choice.
Enterprises tell me that they are hybrid cloud users, while the data in the submission suggests by wording that Amazon is talking about companies who are either all or nothing, cloud-wise. Also, I don’t have enough data on the UK alone to compare with the CMA submission data. I would be very surprised if only 29% of people who were all-in in the cloud with one provider went all-in to the data center. Nobody suggested to me that they had gone from all-in with one provider to either another provider or to multi-cloud. They went hybrid instead. I don’t see this as a problem with Amazon representations, though, but rather to the likelihood that CMA questions didn’t reflect a technical understanding of cloud usage and in particular in the dominance of hybrid cloud.
This interpretive point could create a critical flaw in regulatory view. Yes, it’s true that the data center is competing with the cloud overall, because all hosting options are competitive. However, the real question is when the data center wins, and why, and enterprises have been clear in telling me that the best cloud applications are ones that have considerable workload variability, so sizing a capitalized resource for the important peaks means that on the average a lot of resources would be wasted. Hybrid cloud says keep stuff with dependable workloads in the data center, and cede the highly variable stuff to the more elastic cloud. The cloud and the data center live in parallel for almost all enterprises.
Where does this all-in bias come from? The only companies who talked to me about the relative cost of and difficulty in deployment of a data center versus cloud usage were startups, which happens to be the largest source of Amazon’s cloud business. Largest to Amazon, but not necessarily largest to regulators, at least not in political importance. Enterprises are the big employers and the organizations that impact the most consumers. If startups are to be assessed in any regulatory process it’s critical to call this out and to present the data for enterprises or mid-sized businesses in parallel. The differences between the two groups is too profound to permit mixing them.
Another area where I think the submission could be misleading (again, likely because the submission relates responses to specific points the CMA suggested) is in the area of cloud software. Every major cloud provider supplies a series of software tools via “web service” APIs. These tools can be used in development, and they are similar across providers in broad feature terms but not in API details, so using these tools is very likely to make cloud software non-portable. On the other hand, every cloud provider offers both the ability to host user software in VMs or containers, and to employ tools or suites from platform vendors like Red Hat or VMware. This approach is more generally portable among clouds, and also between cloud and data center. What is true is that some users (few enterprises but likely more mid-sized businesses) are unaware of the distinction here, and also that cloud provider tools, being cloud-specific, are often easier to use than these more portable options, lowering development and maintenance costs. They may also, according to enterprises, be cheaper than one of the portable alternatives.
Like the FTC, the CMA is charged with ensuring competition, so we have to look at the Amazon exchange with the CMA in that light. I think the picture the submission paints is confusing, but there’s a grain of truth here. A cloud provider doesn’t just compete with other cloud providers, they compete with all hosting options. The cloud market would, given that, be competitive even if there was only one cloud provider because the data center would always be another. In another sense, though, the data center is just an example of own versus lease, and for that the data center’s lack of capacity elasticity means it can’t fit some applications that the cloud can fit. You can’t have competition between a suitable and an unsuitable option.
That’s the key point here, and a point that regulators like the CMA seem to miss. The cloud is an agile, tactical, form of hosting. The data center is a static-optimized, secure, form. Only if you cast the cloud as an inevitable destination for all applications does it make sense to say that the data center is always a cloud alternative, and that presumption is fatally flawed. Amazon’s willingness to admit that the data center is a cloud competitor doesn’t mean Amazon is delusional, but indicated that the real truth that the cloud isn’t always the best place to host things isn’t one they’re willing to face at the marketing level. Bodies like the CMA and FTC, if they want to regulate cloud services, have to learn a bit more about how it works.
