Avoiding Unexpected Cloud Economics Pitfalls
Anybody who is considering a move to the Cloud knows that the greatest economic motivation for Cloud Computing is the pay-as-you-go, pay-for-what-you-need utility computing benefit, right? Deal with spikes in demand much more cost-effectively, the public Cloud service providers gush, since we can spread the load over many customers and pass the savings from our economies of scale on to you. The utility benefit is also a central premise of Private Clouds. Build a Private Cloud for your enterprise, the vendors promise, and you can achieve the same economies of scale as Public Clouds without all that risk.
Unfortunately, what sounds too good to be true usually is. There are a number of gotchas on both the Public and Private Cloud provider sides that limit—or even prevent—organizations from obtaining a full measure of the utility benefit. Let’s go back to economics class and take a closer look.
Clouds Like Water?
Turn on the faucet, only instead of water, you get Cloud. Sounds good, but we use water very differently than we do IT resources. With water, we generally use all we need without worrying about price. We may try to economize, and perhaps we’ll go through the trouble of digging a well if we need to fill our pool, but generally we don’t think about the cost of each flush or load of laundry.
The Cloud is just the opposite. The techies might not be thinking in terms of cost, but the bean counters definitely are. For a CIO or purchasing manager comfortable with entering resource costs into annual budget spreadsheets, the unknown nature of the Cloud bill strikes fear into their hearts—and their wallets. Instead of focusing on lowered costs, their worry is increased costs, since Cloud usage is inherently unpredictable. After all, that’s why landlords don’t like including heating costs in the rent. If the tenants aren’t responsible for keeping costs down then pay-as-you-go inevitably translates into pay more—and just how much more is a mystery until the bill arrives.
Enterprise Cloud customers in particular are beginning to push back, and as a result, Public Cloud providers must change their pricing model accordingly. Unfortunately, there aren’t many alternatives to simple pay-as-you go. One increasingly popular alternative that might ease Cloud purchasers’ minds is for providers to offer a tiered pricing system, with a fixed price for any consumption up to a pre-defined threshold, and pay-as-you-go above that. However, tiered pricing is not a panacea. While such a pricing model is straightforward and gives organizations an increased measure of predictability, it still doesn’t solve the problem of cost spikes.
If tiered pricing sounds more like paying for your mobile phone service than for utilities like water or electricity, you’re right. Not only does this approach reduce perceived risks for Cloud purchasers, it’s also a familiar model for the telcos, all of whom are looking to enter the Cloud market, or at the least, grow their existing Cloud offerings. As a result, ZapThink expects tiered pricing to become the norm for Public Cloud services over time, in spite of its drawbacks.
The irony with tiered Cloud pricing is that the more you require elasticity, the greater the risk that you’ll use up your allotted consumption for the month—but elasticity is the most important benefit of the Cloud. Sure, if you have steady, predictable consumption then tiered pricing is low risk, but if all you want is steady, predictable availability, then chances are keeping your resources on-premise or in a traditional hosted facility will be more cost-effective than moving to the Cloud in the first place, since you’re not particularly worried about spikes in demand.
To make matters worse, not everyone likes tiered pricing, of course. Anyone who’s used up their minutes or texts for the month only to be surprised by an excessive phone bill knows what I’m talking about. It seems the mobile phone providers love to play games with their pricing plans for the sole purpose of squeezing every penny out of their hapless customers. I’m sure we don’t want our Cloud providers to play the same dirty tricks.
The Subtleties of Cloud Churn
While it’s a common water cooler pastime to demonize mobile phone companies for their underhanded pricing policies, there is a downside for the providers as well: the dreaded customer churn. Since it’s relatively easy for customers to change phone providers, especially now that number portability is a reality, shifty behavior on the part of providers simply chases away customers.
Cloud churn is a very real problem for Public Cloud providers as well, as the ease of deprovisioning Cloud resources naturally eases the deprovisioning of customers. But there is an extra complication with Cloud churn that doesn’t have a parallel in the mobile phone world: Cloud resources that are no longer being used but still remain allocated to customers. Depending on the provider’s pricing model, the cost to the customer to maintain such resources may be minimal, but it’s not always clear whether those minimal amounts sufficiently cover the providers’ costs.
For example, I get monthly charges on my credit card from Amazon Web Services (AWS) for a few cents each month. I can’t remember how I signed up for AWS, but the amounts are so minimal, it’s not worth my time or trouble to cancel the service. Do those few cents per month cover Amazon’s costs, assuming there are potentially millions of such customers? Perhaps in Amazon’s case—but for less experienced providers with wafer thin margins, the economics might work to their disadvantage.
Furthermore, the proliferation of such idle instances may be a more significant issue for Private Cloud providers, since they typically have constrained budgets for data center buildouts. Amazon may be building new data centers as fast as they can, but your Private Cloud likely has a maximum practical size given your budget for the effort. The last thing you want is to fill it up with idle resources that various people in your organization can’t be bothered to fully deprovision.
The Demotivation Paradox
For the Public Cloud provider, the obvious solution to the problem of idle resources left over from Cloud churn is to charge enough for those resources. Either the cost will motivate people to fully deprovision them, so the argument goes, or at the very least, they generate enough money so that keeping them around is worthwhile for the providers.
But what if we’re talking about Private Clouds here? The way to charge internal customers for using Cloud resources is via chargebacks. And everybody hates chargebacks. Not only are they a bookkeeping hassle, but they also demotivate the consumption of shared resources. We went through this problem when we dealt with shared Services and SOA, and now we’re sharing Cloud resources, but the problem remains: the whole point to the Private Cloud is to achieve economies of scale across the enterprise, but the only way to make such economies work is if most or all divisions participate. Chargebacks, however, discourage that participation.
As it was with shared Services, the way to compensate for chargebacks is through effective governance: establish and enforce Cloud consumption policies that counteract the demotivational effects of chargebacks, and come up with a way to motivate people to follow such policies. While you’re at it, formulate policies governing the deprovisioning of instances that no one needs any more. But in the Cloud, such governance is especially challenging because of the diversity of resources and their corresponding consumption scenarios: policies for provisioning virtual machines as part of IaaS is quite different from, say, provisioning development tools on PaaS. It will take organizations with Private Clouds a good bit of trial and error to get the balance right.
The ZapThink Take
Another downside to the idle-resource-masquerading-as-paying-customer problem is that it makes it very difficult for financial analysts to gauge the health of a Public Cloud provider. This obfuscation can skew traditional metrics like number of customers or revenue per customer, and the distortion may be different from one provider to another. Combine the resulting confusion with the lean profit margins in today’s Cloud space, as providers push their prices ever lower to encourage growth, and you have a recipe for disaster. An ostensibly healthy Cloud provider might suddenly collapse due to a foundation of underperforming customers and idle resources.
Private Clouds face a corresponding problem, as executives review the financials for the Cloud effort. ZapThink predicts a backlash against Private Clouds in the next year or two, as vendors underdeliver on their Cloud promises—not necessarily through any fault of their technology, but rather because the reality of achieving cost advantages with Private Clouds is far more difficult than the vendors’ and analysts’ spreadsheets might have you believe.
If you’d like to learn more about the subtleties of Cloud economics, I’d be happy to have a deeper discussion at Cloud Expo in New York or The Business of Cloud Computing in Dallas, or any of the other conferences I’ll be presenting at. Please drop me a line if you’re interested. I’m curious as to whether issues of Cloud churn or Private Cloud demotivation are concerns in your organization.