More than 90% of organizations will use public cloud services this year, fueled by record cloud computing growth. In fact, public cloud customers will spend more than $50 billion on Infrastructure as a Service (IaaS) from providers like AWS, Azure, and Google. While this growth is due in large part to wider adoption of public cloud services, much of it is also due to growth of infrastructure within existing customers’ accounts. Unfortunately, the growth in spending often exceeds the growth in business. That’s because a huge portion of what companies are spending on cloud is wasted.
Cloud Computing Growth in 2020
While Software as a Service (SaaS) makes up the largest market segment at $116 billion, the fastest growing portion of cloud spend will continue to be Infrastructure as a Service (IaaS), growing 24% year-over-year to reach $50 billion in 2020.
Typically, we find that about ⅔ of enterprise’s average public cloud bill is spent on compute, which means about $33.3 billion this year will be spent on compute resources.
Unfortunately, this portion of a cloud bill is particularly vulnerable to wasted spend.
Growth of Cloud Waste
As cloud computing growth continues and cloud users mature, you might hope that this $50 billion is being put to optimal use. While we do find that cloud customers are more aware of the potential for wasted spending than they were just a few years ago, this does not seem to be correlated with cost optimized infrastructure from the beginning – it’s simply not a default human behavior. We frequently run potential savings reports for companies interested in using ParkMyCloud, to find out whether or not they will benefit from using the product. Invariably, we find wasted spend in these customers’ accounts. For example, one healthcare IT provider was found to be wasting up to $5.24 million annually on their cloud spend, an average of more than $1,000 per resource per year.
Here’s where the total waste is coming from:
Idle resources are VMs and instances being paid for by the hour, minute, or second, that are not actually being used 24×7. Typically, these are non-production resources being used for development, staging, testing, and QA. Based on data collected from our users, about 44% of their compute spend is on non-production resources. Most non-production resources are only used during a 40-hour work week, and do not need to run 24/7. That means that for the other 128 hours of the week (76%), the resources sit idle, but are still paid for.
So, we find the following wasted spend from idle resources:
$33.3 billion in compute spend * 0.44 non-production * 0.76 of week idle = $11 billion wasted on idle cloud resources in 2020.
Another source of wasted cloud spend is overprovisioned infrastructure — that is, paying for resources are larger in capacity than needed. That means you’re paying for resource capacity you’re rarely, or never, using.
About 40% of instances are sized at least one size larger than needed for their workloads. Just by reducing an instance by one size, the cost is reduced by 50%. Downsizing by two sizes saves 75%.
The data we see in ParkMyCloud’s users’ infrastructure confirms this, and in the problem may be even larger. Infrastructure managed in our platform has an average CPU utilization of 4.9%. Of course, this could be skewed by the fact that resources managed in ParkMyCloud are more commonly for non-production resources. However, it still paints a picture of gross underutilization, ripe for rightsizing and optimization.
If we take a conservative estimate of 40% of resources oversized by just one size, we find the following:
$33 billion in compute spend * 0.4 oversized * 0.5 overspend per oversized resource = $6.6 billion wasted on oversized resources in 2020.
The Extent of Wasted Cloud Spend
Between idle and overprovisioned resources alone, that’s $17.6 billion in cloud spend that will be completely wasted this year. And the potential is even higher. Other sources of waste include orphaned volumes, inefficient containerization, underutilized databases, instances running on legacy resource types, unused reserved instances, and more. Some of these result in significant one-off savings (such as deleting unattached volumes and old snapshots) whereas others can deliver regular monthly savings.
That’s a minimum of about $5 million wasted per day, every day this year, that could be reallocated toward other areas of the business.