With the global economy in meltdown, you might think it’s time to hunker down with your private data centre, bringing workloads back from the cloud (“repatriation”) to save money.
You’ve heard the marketing spiel: Cloud is great but it’s even better with a hefty dose of private data centre investments because, um, some choose to “more shrewdly optimise their applications and expenses by matching the right application with the right environment” (a little Dell sophistry for you).
They point to this report or that survey that finds some large percentage of CIOs plan to repatriate workloads due to cost, security, etc.
I’m sure your server vendor wants you to believe this. Sadly (for them), it’s not true.
Even if it were true that many enterprises plan to repatriate workloads, you shouldn’t be one of them. There are many reasons you and virtually every other enterprise are moving workloads to the cloud, reasons that have only become more pronounced as macro-economic conditions have worsened.
Option value and the cloud
Of course, this counsel might not even be necessary. After all, even if you want to buy more servers, you likely can’t — not as fast as you might like, anyway. Gartner analyst Lydia Leong points out that “enterprises [are] moving to cloud because data supply chain issues are making it hard to expand their on-premises footprint.”
This is just as well because you really don’t want to bulk up on servers, especially not when it’s so hard to predict customer demand.
You may have read the recent diatribe from 37signals co-founder David Heinemeier Hansson (DHH), “Why we’re leaving the cloud.” You might also have found yourself nodding when he summarises, “Renting computers is (mostly) a bad deal for medium-sized companies like ours with stable growth.”
Except that you almost certainly don’t have stable growth. Few do. We’ve been living in the midst of “uncertain times” for years, and it’s not getting any more certain.
As Stedi CEO Zack Kanter calls out, “Imagine if you committed to buying repatriated infrastructure based on 2021’s projected growth rates.” Bad, right? He goes on, quoting Toyota luminary Taiichi Ohno: “A machine becomes fantastically expensive when we fail to sell the expected number of products.”
Sure, it’s possible that if you happen to have an application that basically never changes, or that grows (or declines) very predictably, it might make sense to optimise this by operating it on your own server. For me, however, such applications don’t sound particularly interesting.
They’re not bet-your-business-on-the-future applications. They sound more like an application graveyard. The cloud, by contrast, enables an enterprise to scale up or down according to demand, saving (or spending) money in line with the application’s success. This sounds like a much better model for enterprises hoping to optimise costs in an unpredictable economy.
Tuning the cloud
I’d write a point-by-point takedown of DHH’s arguments against the cloud (and use them as a proxy for reasons you might be thinking of repatriation yourself), but I don’t have to. Geoffrey Greene of GNDS Consulting has already done so, and it makes for compelling reading.
For example, you might think, as DHH apparently does, that the cloud is only good for “highly irregular” workloads with “wild swings or towering peaks in usage.”
Yes, the cloud is good for such workloads, but it’s also great for staid workloads, as Greene indicates: “If you do have predictable workloads, AWS [and other cloud providers, for that matter] provides the concept of ‘reserved instances’ that can give up to a 70 per cent discount.” He then goes on to suggest “It’s very easy to architect an expensive app on AWS” in ways to make it cost-effective.
In fact, Greene is probably right that DHH’s arguments for repatriation and against cloud are mostly vacuous, unsupported by facts. He concludes that maybe DHH and 37signals should “admit we just want to do this because it’s gonna be cool.” Not because it’s better or cheaper or faster, but because they want to geek out with their own hardware and software.
Good for them. But taking this road will very likely not be good for you. Let’s face it: You’re not going to be able to out-cloud the cloud providers. They’re better at security, maximising performance and efficiency, etc.
They simply are. You employ great people, but those same people will be much better off focusing on putting higher-value cloud services to work for you, rather than mucking about with servers for storage, compute, etc.
Sorry to pick on Dell, but the company seems to be hell-bent on this cloud repatriation message. Dell has taken out paid posts to call repatriation a way to “modernise IT without trade-offs.”
Repatriation requires the biggest trade-off of all, asking an enterprise to keep dumping money into infrastructure that is costly, hard to manage, and not going to get you closer to the digital transformation required to help you care for customers.