When businesses overspend in the cloud, the most commonly-recommended cloud cost reduction steps are to rightsize the assets you use, terminate the assets you don’t use, and take advantage of committed use discounts when you can. Unfortunately, these cloud cost reduction steps don’t work for all the people all the time.
Most businesses spend more than they need to in the cloud. The extent to which they overspend is a matter of some conjecture because most statistics related to cloud spend are produced by vendors of cloud cost reduction solutions. These statistics are invariably based on the vendors’ customers that have come to them looking for help reducing their cloud spend.
Where statistics are produced by independent research companies, they tend to be vague or based on small or specific sample sets. Consequently, statements like “organizations have frequently noticed public cloud bills that are two to three times higher than expectations” don’t make a lot of sense unless you know what the organizations’ “expectations” were.
Why Businesses Overspend in the Cloud
One of the main reasons businesses overspend in the cloud is because they are “operationalized” for on-premises infrastructures. This means they’re geared for the “sunk” capital expenditure costs of an on-premises infrastructure rather than the self-provisioning nature of cloud computing. Whereas in an on-premises infrastructure, operating expenses are negligible; in the cloud operating expenses are all there is, and this problem is exaggerated by myth “you only pay for what you use in the cloud”.
The myth is a victory for marketing inasmuch as cloud service providers have been using it to promote their services since the dawn of cloud computing. Had cloud computing been marketed as “you only pay for what you provision in the cloud,” businesses might have been more careful about what they provisioned and what they did with their provisioned assets once they finished using them. As it is, some cloud users have a mindset that it’s okay to over-provision and underutilize—until the bills arrive.
The Most Commonly-Recommended Cloud Cost Reduction Steps
When the bills arrive, the pressure is on to reduce cloud spend. What’s the first thing you do when you need to know how to do something? You consult the Internet, where the most commonly recommended cloud cost reduction steps are to rightsize the assets you use, terminate the assets you don´t use, and take advantage of committed use discounts when you can.
This is fine if you have complete visibility of your cloud environment and access to utilization metrics from which you can determine which assets are underused or unused, and which have steady and predictable workloads that are suitable for committed use discounts. If you don’t have complete visibility of your environment and access to utilization metrics, minimizing costs will be impossible.
Why Some Cloud Cost Reduction Steps Don’t Always Work
Even with the appropriate tools to see and analyze assets, some cloud cost reduction steps don’t always work for all the people all the time. Take rightsizing VM instances, for example. Most cloud service providers (Google is the exception) have fixed sizes to choose from that double in capacity with every increase in size and half in capacity with every decrease in size.
Therefore, if you have deployed a General Purpose VM instance with (for example) 2 vCPUs and 8GiB of storage, and it has peak utilization of 60%, it’s not impossible to downgrade the VM instance to a smaller size. What it may be possible to do depending on the individual metrics is move the workload to a different family (i.e. compute optimized or memory optimized) to reduce spend that way. However, the work involved in finding the appropriate option can cost more over time than what is saved in operational costs.
Can You See all the Zombies in Your Cloud Environment?
In some cases, even with the visibility tools provided by cloud service providers, it may not be possible to see all your unused “zombie” assets. Users have commented that it’s very difficult to find unattached Elastic IP addresses in the AWS Console or obsolete snapshots in the Azure Console. Google Cloud’s Asset Inventory tool is still in beta and doesn’t yet support all asset types.
The problem with this is that if you’re unaware of all the components automatically attached to VM instances and other assets (storage volumes, load balancers. IP addresses, snapshots, etc.), you won’t know whether there are zombies in your cloud environment or not. There’s a saying in cloud computing that “you can’t manage what you can’t see,” so it’s vital that you have complete visibility of your cloud in order to terminate, delete, and release any assets you aren’t using but are still paying for.
Committed Use Discounts have to be Fully Utilized
Taking advantage of committed use discounts (i.e. AWS Reserved Instances) is a common sense measure to achieve a cloud cost reduction, and many businesses make full use of these discounts—initially. A common issue with committed use discounts is that businesses often treat them as “set and forget” purchases and fail to manage the purchases during their lifecycles to ensure they’re fully utilized.
It´s all very well paying three years upfront to achieve a 50% discount. But if the utilization of the VM instances assigned to the reservation changes, you could be using only 50% of the reservation while running other VM instances at On Demand prices. That would effectively mean you’re paying upfront for no discount at all. So, inasmuch as purchasing committed use discounts is a cloud cost reduction measure that does work, the purchase has to be properly managed to ensure it continues working.
CloudHealth: A Solution to Overcome these Issues
The CloudHealth cloud management platform is an ideal solution to overcome these issues and help businesses achieve a permanent cloud cost reduction. CloudHealth consolidates data from all your assets in the cloud—whether your business operates in a single cloud, multiple clouds, or in a hybrid cloud environment—and analyzes asset utilization to make recommendations for matching assets to their workloads (i.e. upgrading, downgrading, or moving workloads to a different asset family).
CloudHealth will alert you to underused and unused zombie assets you can terminate, delete, and release so you don’t have to go looking for them, and also notify you if committed use discounts are underutilized.