- Micah Wheat
Like hairlines, waistlines and the federal budget, cloud costs seem to only move in one direction.
But are cloud infrastructure costs necessarily a positive, monotonic function of time? Of course not.
As cloud costs have ballooned, people have come up with increasingly clever ways to manage them. Here are the three main strategies for reducing cloud costs in 2023.
They are not mutually exclusive, and some companies will find that it makes sense to use two or three of these strategies in conjunction.
Strategy #1: Optimize reserved instances
The “big three” cloud providers employ what we call “crack dealer pricing.” They offer very generous credits to newborn startups, often worth $100,000 or more, making the “first taste free.” But these credits typically expire after a year, and, in that time, bad habits often get formed.
Fortunately, the remedy is simple: move as many workloads as possible over to reserved instances. Given that reserved instances can be more than 70% cheaper than autoscaling, even radical over-provisioning can yield major savings.
A “quick and dirty” first step that can nonetheless yield savings in the ballpark of 40-50% is to provision double the resources you expect to need.
Another option for engineering leaders looking to squeeze even more juice out of reserved instances—and to have this done automatically—is to use one of a plethora of SaaS tools like Cloudability, Vantage, GorillaStack, Usage.ai, CloudBolt or nOps that handle reserved instance "rightsizing" behind the scenes.
Strategy #2: Leverage collective bargaining
A second cloud cost cutting strategy that is becoming increasingly popular is to get a discount via “group buying,” kind of like what PEOs do for insurance.
These vendors claim to offer more elasticity than reserved instances at a price lower than what you’d get from autoscaling.
Strategy #3: Improve cloud efficiency
The primary way to reduce cloud costs in the long run will always be to consume fewer cloud resources. Engineering leaders estimate that around 30% of their cloud spend is wasted on average, so most companies have bushels of low-hanging fruit ripe for the picking. (If this figure is accurate, firms will waste around $180 billion in cloud spend this year.)
One way to identify inefficiencies is to use a tool like AWS Cloud Watch, Datadog Cloud Cost Management or Ternary to see which cloud resources, e.g. databases, storage buckets, compute instances, and tags, are racking up the highest costs. This is a helpful first step for companies that have no idea where their costs are coming from.
A limitation of these solutions is that, to the extent that products, features, teams and customers share cloud resources, it can be near impossible to pinpoint the culprit. Companies that need visibility into their costs at the sub-resource level can either build custom observability tooling or use Dashdive.