I know what you’re thinking.
Fool! What do you mean, the cloud won’t save me money? Of course, it will. Just think of all the ways cloud is cheaper / faster / better. With cloud, you get rid of:
- Data center facilities and personnel
- Server, storage, and networking hardware purchases
- Systems installation, break/fix, patches, and upgrades
- Hardware support contracts
- Hardware refresh cycles
- Scheduled infrastructure maintenance windows
- Hardware, software, and firmware compatibility issues
- Capacity management guesswork
Cloud is simple, agile, elegant, and pay-as-you-grow. And, everyone is moving to it. Gartner says so:
“By 2025, 80% of enterprises will have shut down their traditional data center, versus 10% today.”
To which I reply:
All that may be true, but if your primary motivation for moving to cloud is to save money, you will be disappointed. Here are five reasons why.
1. Jevons Paradox
I’ll admit, I’d never heard of Jevon’s paradox (or its close but more colorful-sounding cousin, the Khazzoom-Brooks postulate) until perhaps six months ago. But I’ve heard it many times since. The idea dates to the mid-1800s and an English economist, William Stanley Jevons. He was trying to explain why, when the cost of coal dropped, its consumption soared. Anyone who has taken a basic economics class within the last hundred years or so would recognize this as a manifestation of the demand curve. But the idea was new in Jevons’ time.
Within the context of cloud computing, Jevons Paradox implies that as IT resource consumption gets easier, demand will increase. This is what cloud does. According to the U.S. National Institute of Standards and Technology, two key characteristics of cloud computing are On-Demand Self Service and Rapid Elasticity. Because the cloud makes IT consumption easier, people will use more of it.
Well, duh! Do we really need a pseudo-scientific sounding term to explain that?
Apparently, we do. The increased popularity of the term seems to be an attempt to justify why overall IT spending goes up when organizations move to cloud.
CFO: “Why are our cloud costs growing out of control?”
IT Director: “It’s not my fault. It’s Jevons Paradox!”
2. Shadow IT
In the on-premises data center world, IT expenditures and capacity are typically determined once a year as part of the annual budgeting cycle. In preparation, the CIO reaches out to her various constituencies and asks them to estimate the requirements for the next 12 months. If they reply at all, department heads usually come back with projections of little-to-no growth for the coming 12 months. These estimates are then enshrined in the annual budget.
At some point in the budgetary year, department heads and line of business owners ask IT for more than they had estimated. IT replies that there aren’t sufficient funds in the budget for their request. The business responds by finding ways to work around corporate IT, resulting in so-called shadow IT. Of course, most business units don’t have the resources to run their own data center operations outside the view and control of corporate IT. But because cloud computing makes IT consumption simple and fast, it facilitates shadow IT. All that’s required is a credit card.
Once all the islands of IT are discovered and the costs added up across the organization, guess what? The bill is higher than anyone expected or budgeted for.
3. Shared Lunch Syndrome
Let’s return for a moment to our discussion of on-premises data centers. They typically contain many shared resources, things like:
- Data center facilities and personnel
- Core switches and routers
- Enterprise storage arrays
- Shared server hardware
While IT would love to implement chargeback to consumers of these resources, the challenge is nearly impossible. (Who, for example, would pay for the hot-spare disks in the storage array or the unused ports on the switches?) As a result, the costs for running IT are typically allocated in the form of a corporate tax on the rest of the business. Each department pays their share, often based on organization headcount.
I call this the shared lunch syndrome. The concept goes like this: a bunch of us are going out to lunch, and we’re going to split the check evenly. In the absence of individual chargeback, showback, or (my favorite) shameback, what is your incentive to purchase an inexpensive meal? (It’s a trick question. You have none.)
Years of shared lunch syndrome have inured business heads to eschew frugality and to over-consume — in effect, to buy the most expensive meals on the menu. When cloud is introduced this habit is exacerbated, and costs increase.
4. Infrastructure as Code
To a user asking for services, corporate IT can seem slow and bureaucratic. Take, for instance, the process flow of a typical systems request.
Cloud, on the other hand, is designed to be agile. Requesting and instantiating resources is as simple as writing a few lines of code:
resource “aws_instance” “cryptominer” {
ami = “ami-12345678”
instance_type = “p3dn.24xlarge”
tags {
Name = “FreeMoney”
}
security_groups = [ “${aws_security_group.my_security_group.id}” ]
}Within seconds, your shiny new virtual machine is up and running. (Beware Jevons Paradox!)
5. Lack of Oversight
To paraphrase the character Syndrome from The Incredibles, when everybody is responsible for managing their spending, nobody will be. In the old IT world, the IT department controlled IT spending — and was resented for it (see point 2 above). Yet when spending decisions are pushed down to a line of business managers, requests often receive rubber-stamp approvals and spending grows unchecked. An absence of organization-wide visibility and control, coupled with cloud’s ease of consumption, is a perfect recipe for The Scary Bill.
The variable and dynamic nature of cloud consumption require constant vigilance. It isn’t enough to simply review your cloud bill each month. Take, for example, that virtual machine you created in point 4 above. What if you didn’t create it, but someone else gained access to your account and is running jobs on your tab? The hourly on-demand price of a p3dn.24xlarge instance is $31.218. If used constantly, that’s $749.23 per day or $22.8K per month. If they’ve spun up, say, 20 instances in a distant region, you could be out over $450K before you even see your next monthly cloud bill.
That’s another thing cloud makes possible. In an on-premises world, your exposure is limited by the available capacity of your data center. In the cloud, your liability is virtually unlimited. For folks who dabble in the stock market, an analogy is stock options versus futures. With options, the most you can lose is what you own. With futures, you can lose far, far more.
There is Hope
If, at this point, you’re feeling depressed because I’ve convinced you cloud savings are a chimera, take heart! As the use of cloud has grown and matured, so too has the recognition of a need for financial discipline in its management.
In much the same way that DevOps applies a developer perspective to IT operations, the FinOps Foundation was recently formed to apply financial rigor to cloud. FinOps is a combination of tools, rules, roles, and responsibilities focused on cloud cost management and optimization.
