As cloud reaches more mature adoption levels in many organizations, those at the leading edge are starting to hit the same issues we’ve seen with other technologies. Early adoption is driven by a major change in value, so the cost side of things often gets less attention.
As processes mature, well run companies start to focus on costs, not to strangle innovation, but to ensure that discipline starts to be baked into operations lest they spiral out of control and require much more painful interventions when tougher times arrive. Building for the long-term requires building in the right structures before they become an “I wish we’d done this earlier” scenario.
Dealing with failed experiments is easy; you just turn them off. But what of the successes? “Once a company has success, then they need to get more disciplined,” says Densify CEO Gerry Smith. Discipline helps you to scale your successes so that they remain successes, and that means cutting waste while investing in value. Cloud services like AWS are cheap to fail in, but expensive to succeed.
“It’s not a problem until it’s a problem,” Smith says, “And then it’s a huge problem.”
“One software-as-a-service company was spending $24 million on cloud a year, on $150 million in revenue,” he said. 16% of gross margin is a lot to be spending without at least some scrutiny on whether it could be done more efficiently.
“Analytics can help us to optimise,” says Smith. “You can start optimizing earlier if it’s easier to do.”
Let us be clear: premature optimization is a fool’s errand and many promising ideas are killed in their infancy by overly restrictive CFOs. We want our experimental failures to be cheap so that we can perform more experiments and learn new things. But over-spending on our successes wastes money that could be invested in adding new features, growing the sales team, or scaling up customer support so that growth doesn’t outpace our ability to manage it.
“Application owners are always happy if the performance is good,” says Smith, “While finance people want to save money, but often don’t know the lingo of the application teams.” The communications mismatch can result in local optimizations of spending that hurt performance, or stellar performance that costs vast amounts more than it needs to. Burning cash is easy when capital is easy, but if times change and finance gets tight, the lack of discipline can kill an otherwise promising company.
Optimization problems get harder as the number of moving parts increases. As the combination of options from cloud providers grows—the array of options available from AWS alone is staggering, and constantly expanding—the optimization complexity soon grows beyond unassisted human comprehension. How do you know that the choices you made six months ago are still optimal when a dozen new instance types are now available? Or that these specific changes are the best ones to make? Will this hurt performance? Will you still be over-spending?
“Picking the right instance is better for both performance and spend,” says Smith. “Our value proposition is in helping people to have confidence they’re choosing wisely.”
Optimization problems are not new. Logistics, transport, and manufacturing companies have been dealing with them for decades as supply chains became ever more complex. We’re now seeing the same sort of industrialization of software and IT infrastructure supply chains, so a growing ecosystem of tools and support structures should come as no surprise.
Densify is just one of the many tools we should expect to see in a maturing industrial setting that needs to operate at scale. We should be more surprised that it has taken us this long.
This article first appeared in Forbes.com here.