What’s Driving Your Over-Provisioning Problem: Technology, Pricing, or Personality?
Park Place Hardware Maintenance
IT professionals and business units need to shift perspectives to one in which sufficiency is expected and it’s the overages that must be managed.
One-third of your cloud spend is likely going to waste. That’s the result from RightScale’s State of the Cloud Report for 2018 (available here with free registration). Information Age has similar bad news. According to its surveys, companies are spending two to three times what they expected in the cloud.
The industry is finding that much of the sprawl data center managers worried about on their premises wasn’t eliminated, it simply moved. That’s a shame, because the cloud was supposed to address precisely this problem using elastic, self-provisioned, on-demand, pay-as-you-grow services.
So what’s at the root of IT’s continued overprovisioning? Actually, there are a few possibilities.
The Scarcity Mindset
There is a psychological response to provisioning, and it’s entrenched in the enterprise mindset from the pre-cloud days. When we had to look forward months or years to plan data center build-outs, ensuring enough capacity to last until the next big expansion was paramount.
It’s not so different than growing food for the winter—having too much is considerably less of a problem than having too little. Thus the annual zucchini overproduction leading neighbors to beat down the door begging you to please, take some squash.
We are hardwired to avoid scarcity, but the instinct becomes less adaptive in the cloud—much like humans’ ingrained love of fat and sugar isn’t doing us any favors now that there are 24-hour fast food joints.
IT professionals and business units need to shift perspectives to one in which sufficiency is expected and it’s the overages that must be managed. Think calorie counting. Rather than stocking up on capacity “just in case,” it’s often better to provision with an eye toward average loads and then consider how to deal with peaks elastically. Rather than transitioning the same set-up from an in-house data center—with all its underutilization issues coming along for the pay-by-the-minute ride—a complete rethink is generally necessary when moving to the cloud.
This will require training, as anyone involved in self-provisioning, whether a developer or a marketing director, can fall prey to the overprovisioning instinct.
Complex Pricing Models
So teach people not to buy extra cloud services. Sounds simple, right? Unfortunately, it’s only the first step.
Next, those same individuals need to parse the cloud offerings. There are literally tens of thousands of prices for VMs (virtual machines) on the three leading cloud providers alone. As CIO Magazine discovered:
AWS lists nearly 150 “products” grouped under 20 different categories (e.g. compute, storage, database, developer tools, analytics, artificial intelligence, etc.). That portfolio makes for well over 1 million different potential service configurations.
Informed purchasers need to navigate all those choices and then compare them with Google Cloud, Microsoft Azure, and smaller providers to make sure they’re getting the right capabilities and the right deal—yeah, it’s not so easy.
The task can be especially difficult for less technical personnel, and they are likely to default back to the “just in case” provisioning model. With pricing often given in pennies, the actual costs over the months or years a workload or application may run are not always readily apparent so they don’t see much downside in buying a little extra, just to be sure.
Technology Barriers to Full Transparency
Of course, IT pros want a technical solution to all of this. Simply tell us what we’re spending, where, and for what and compare it to what we’re actually using—and we’ll fix it. Better yet, we’ll automate it so it fixes itself, no matter what the lines of business are doing.
There’s no doubt, such solutions are on the way, but for now the problem is more complex than we generally credit. It is difficult to peer down through the layers to truly understand utilization and make the right provisioning decisions, such as whether to add CPU or memory instances.
The bottom line, provisioning correctly means matching supply and demand across resources in real time, and the tools to provide that level of transparency are still a work in progress. That doesn’t mean IT pros should give up, but they should expect to devote significant time, attention, and technology to getting their cloud provisioning right. Fortunately, the cost-savings are usually there to justify the investment.