Why it is so hard to optimize cloud costs for your containerized applications

This is called cloud waste, and it’s a real problem.

Executives surveyed in Flexera’s 2020 State of the Cloud Report estimate that at least 30% of their cloud spending is wasted, and they go over budget by 23%, on average. The 2019 European Insight Intelligent Technology Index (IITI) report showed that 30% of cloud spending was bankrolling services lying dormant, resulting in a massive annual wasted cloud expenditure. In the same report, 39% of the answers mention the difficulty to plan and allocate budget for cloud consumption.

From 2019 European Insight Intelligent Technology Index (IITI)

At the same time, businesses aim to increase their spending on cloud services due to COVID-19. In the IITI report, 36% of enterprises cited the lack of visibility as the primary reason behind cloud waste. That’s why tools like CAST AI can help by revealing the true performance of various machines offered on the market and enabling businesses to manage their cloud costs efficiently.

So, why is controlling cloud costs so difficult?

Problem 1: you only know the cloud cost at the end of the month

Organizations that use Infrastructure as a Service (IaaS) cloud platforms have no way of predicting budgets per project in advance. They also stand no chance at understanding when their resource usage creeps up to potentially put projects over budget. About 40% of instances are sized at least one size larger than needed for business workloads.

Now, you can always set hard limits per project to make sure that you don’t go over budget. But how can you know what this limit should be?

Despite what cloud providers suggest, the old saying stays true: You only know how much your cloud bill is on the last day of the month. By then, there’s nothing you can do to reduce it.

Problem 2: It is hard to remove resources after the fact

When getting a cloud platform, you have no assurance at all that the IaaS resources are being used in the most optimal way. You can’t know whether the VMs and other resources you requested and purchased are the best choices for your workloads.

Once your application is in production, you can easily measure excess capacity. But the same isn’t true for removing that excess capacity from your system.

Problem 3: Which resources to select in the first place?

A CPU core from one cloud provider isn’t the same as from another cloud provider, today and tomorrow. Similar resources come at different prices and with different capacities across practically all of the major cloud service providers. That’s why choosing the best cloud platform is so challenging.

You will always find another VM shape of similar capabilities that costs less than yours. Seriously, there’s always another option waiting around somewhere. And that’s only today, just think about tomorrow!

Look at this example: these two machines are similar on paper, they both have 4 cores. Oracle VM has more memory, but the price is almost the same. However, our performance benchmark shows that, in fact, OracleVM.Standard.E2.4 brings almost twice the blended compute capacity over 24 hours than the Google Cloud equivalent.

CAST AI Cloud Compare tool, June 2020

Problem 4: It gets cheaper if you have 2 minutes

Major providers include alternative pricing options apart from the standard “pay as you go” (PAYGO) mode.

While PAYGO is the standard retail published price for resources, other options might be interesting for those looking to cut costs. Preemptible instance pricing (sometimes referred to as spot instance pricing) is one of them.

However, preemptible instances are hard to use for most customers because they can be interrupted at any time or have a time limit. AWS delivers a notice 2 minutes prior to interruptions. Other cloud providers offer something similar.

For example, 4-core C5.XLARGE cost is a 4-core VM on AWS. The regular price is $0.172 per hour (or about $124.10 per month).

CAST AI Cloud Compare tool, June 2020

The same VM on the AWS spot market costs $0.0405 per hour, a 76% discount over the regular price.

AWS spot pricing for c5.xlarge in June 2020

Since most businesses don’t have automated processes to deal with such interruptions, they can’t benefit from superior pricing – which, in our experience, can often yield savings of as much as 90%.

Problem 5: Vendor lock-in

Once you choose a cloud service from AWS, you’re stuck with AWS for the rest of your application’s life. Cloud providers use many methods to prevent engineers from moving their workloads to other, cheaper cloud providers. This is especially true in situations where you use the native services of the cloud provider.

This is called vendor lock-in. Stopping the use of one cloud platform and changing it to another requires an intolerable, painful, and boring application refactoring, with an unreliable ROI at best. Multi cloud is not an option.

But there’s still hope… You can now ace cost control using CAST AI

At CAST AI, we want to empower developers to use whatever cloud providers their containerized application needs, in any region they like.

We believe that moving workloads from one cloud to another should be a trivial and automated task so that businesses can benefit from cost optimizations and have greater control over their cloud budgets.

To accomplish that, organizations need to gain more visibility of their cloud services to understand their usage better and pick the best match for their requirements.

Example of a Cluster that stretches across three clouds:

By using tools such as CAST AI, you can gain visibility into the performance of your cloud platforms to control its costs better and never go over budget again.

If you’re curious about how your cloud performs, try our free Cloud Compare tool. Leave us your email address to receive updates and gain the opportunity to become one of the first users of our full product.

Please send us an email to [email protected] for an invite to our beta. Or use the chat in the bottom right corner to get in touch with us.