Remember the old saying, “Content is king”? Remember when tech authors, including me, played on the phrase “video killed the radio star” to demonstrate how network service was justified by what it delivered? Well, SDxCentral may not realize it, but they’re asking the same sort of question about data center capacity in this piece, and that the answer is the same too.
The meat of the story is that a research firm had reported that the hyperscalers, by which they mean public cloud providers, had 41% of worldwide data center capacity, while enterprises had only 37%, down from 60% only six years ago. They attribute this to the shift of enterprises to the cloud, saying that in 2012 enterprises spent twelve times as much on data center hosting as on cloud services, where today they spend three times more on cloud services than on the data center.
I have to say first of all that only 11 of 341 enterprises who offered me comments on cloud services versus data center spending even said they spent more on the cloud, and none said their cloud spending was triple their data center spending. It’s possible that the “spending” here is new capital spending in the latter cases versus cloud expense in the first, which isn’t a useful comparison, but there’s a deeper point here too. How much of cloud spending is really an erosion of the data center versus accommodating a shift, the shift that “content” has been all about from the first?
The biggest driver to cloud service growth is the Internet, not a flight from the data center. Interestingly, 4 of the 11 enterprises who spent more on the cloud than on data center infrastructure said this was because of a shift to an online-centric business model. That’s also by far the largest reason why enterprises say they use the cloud today. You only have to read all the stories about the runaway cost of the cloud (or hear, as I do, the same tales from enterprises) to understand that for traditional applications, the cloud is not more economical. What’s going there isn’t old stuff moving to a lower cost, but new stuff, new “content”.
My own use of the Internet is focused not on streaming video/audio content, but on visiting company websites. Since I do use streaming for both my audio and video needs, I’d admit that most of my actual traffic is probably content of that sort, but in terms of activity, meaning the number of times I do something, visiting company websites dominates. When I’m working, it’s almost my exclusive use broadband. Over the last couple of decades, companies have shifted their marketing and sales increasingly to optimize their online relationships. It’s this online focus that has magnified cloud usage, so it’s really online sales/marketing content delivery that’s driving enterprise spending on hosting.
Of my 341 enterprises, 174 do admit that they have “slimmed down” some data center applications. They say that, of course, any elements of those applications that relate to a GUI have been minimized in favor of new cloud front-ends, not so much as a deliberate and proactive shift but because the “onlineification” of their sales, marketing, and support activities have taken over the functions these data center elements used to perform.
Enterprises tend to regard their data centers as exactly what the name suggests, a “center for data.” That means that databases, and the transaction processing related directly to them, is the core piece of the data center mission. Old-timers in the CIO organization will even point out that the Internet, via web servers, was already eroding the GUI piece of applications before the cloud came along, and that their use of the cloud was in some ways an expansion of the web server and not a migration of something from the data center.
The reason this is important and not just an exercise in statistical precision is that the future of the cloud and the data center really both depend on online behavior more than on relative cost or hosting efficiency. In some ways that’s a good thing, because it means that the current flap over cloud costs may not be critical for the future of the cloud. If the cloud is really driven by online appetites, then it seems a safe bet that it will continue to be important. In some ways, cloud aficionados may see it as a bad thing because the need for more cloud to improve businesses’ online presence is difficult to quantify, and perhaps difficult for pundits to pontificate on. “Everything is moving to the cloud” is easier.
Orderly evolution of the cloud-versus-data-center balance might be upset from the outside, though. One way is truly “outside”, meaning related to the edge application hosting issue. The other is AI.
We can argue that “online” has moved IT outside, and that implicates the whole issue of edge computing, both as a contributor to the value proposition for the cloud and for the data center. Edge computing is where enterprises host applications that support real-time, real-world, control. As I’ve been saying for a decade, the next frontier of IT is to bring both worker empowerment and consumer sales, marketing, and support, out into the real world. This means harnessing IoT to understand what the real world looks like and how people move, live, and work in it. Today, the amount of edge computing that’s done anywhere other than on premises, close to the processes it supports, is statistically insignificant. Not only that, enterprises say that over 80% of all data collected by edge sensors doesn’t go back to either the cloud or the data center, even as a record of things. If we were to push “online” more into a real-world-and-real-time framework, if it became important for review and analysis, where would this data then likely be collected and stored?
Then there’s AI. The article suggests that AI will further tip the balance, and it will, but not in the direction the article proposes. AI in the cloud, according to enterprises, is not able to deliver the business value they want without compromising data sovereignty. Why would enterprises, who have almost totally rejected the idea of storing business-critical data in the cloud, send it to cloud AI? The largest area of opportunity for AI growth demands hosting in-house.
It’s hard to justify how views that seem totally at odds with what enterprises tell me somehow percolate. I think it’s a sad but common problem of survey bias, meaning that surveys are often wrong because they target the wrong people in a company, the respondents want to look good and so say what they think is most appropriate, or the survey questions tend to elicit a specific answer. In any event, what I’m saying here is what enterprises, unprompted, are telling me, and I have to say it makes sense. Let me know if you believe otherwise.
