- The Trade-offs of Multitenancy
- CPU Performance, Meet the Internet
- Paying Too Much for Cloud Compute? Here's Why
- Picking the Right Operating Systems
- Picking the Right Memory Configurations
- The Concept of Reserved Instances
- Going Off-brand
- Call to Action
This chapter is from the book
This chapter is from the book
This chapter is from the book
Paying Too Much for Cloud Compute? Here’s Why
A common theme in this book is that most who leverage public clouds pay too much. You’ll find study upon study that proves this statement correct, but what’s at the heart of the issues that break cloud budgets? If we just focus on compute issues for the moment, a few issues seem to tip the scales.
As covered previously, most of those who configure and allocate cloud computing compute resources do so without a good understanding of what the application does and how it does it. When the choices are based on uneducated guesses, most of those tasked with picking a compute configuration will select more resources than the application needs. Or, the budget-conscious could pick too few resources and then end up with an application that fails as it runs out of CPU horsepower or perhaps memory. Overbuying typically won’t be identified as a problem until cloud costs finally come under audit. Eighty percent of the applications I see have typically overpurchased compute resources, spending as much as three to four times the money they should have on resources that won’t be fully utilized.
Another issue? The cost tracking that cloud providers make available are not set up to let you know when or if you’re overspending. To their credit, most cloud providers do offer tools to determine the size of the resources that you need. However, the tools have limited value. Typically, they’re just questionnaires that those about to migrate and build net-new applications will need to fill out. In many cases, the new clients don’t know what to tell the questionnaire.
Many of the mistakes being made today are by IT leaders who are just learning to understand cloud compute. How eager would you be to point out your own mistakes, ones that could be costing the business millions of dollars in unnecessary cloud resources?
The solution to this problem is obvious, but one that’s rarely considered due to cost concerns. The better option is to implement a complete FinOps (Financial Operations) system that includes people, processes, and tooling that can automate the ability to determine when and if you’re spending more cloud dollars than needed.
A FinOps system should also suggest changes that could reduce spending without impacting application performance and reliability. A FinOps approach and tooling will pay for itself compared to the costs involved to do a migration wrong the first time and then go back and fix it later. Implementing a FinOps system can provide a larger and ongoing return on its initial investment—this considering that ongoing processes and automation allow us to monitor costs, create policies around managing costs, and optimize the use of cloud resources with costs in mind.
Here are a few more ways to ensure that your cloud compute costs are more in line:
Accept outside help. Hire people who can provide you with an objective opinion as to what you’re spending on compute and where it’s in line with what the workloads need. Consultants often provide this service for several businesses and thus have experience in what’s optimized and what’s not.
Deploy FinOps. This is a subject that we cover heavily in this book. Basically, it’s the ability to track cloud costs by usage, manage cloud negotiations to obtain the best prices, and do cloud cost forecasts to understand what’s spent now and what will be spent in the near and far future.
Gain a wider understanding of cloud costs. In many instances, the focus is on a single cloud provider. Rather than just work within their walled garden, it’s a good idea to understand what the other cloud providers offer in terms of costs and compute configurations. You may find that you can pay half the cost for the same configuration on one provider versus another. We cover multicloud later as it relates to the consideration of compute configurations from other providers.
Keep in mind this all goes to the benefits of cloud cost optimization. This is both an approach as well as sets of software that allow us to optimize costs using automated systems. This means that we really don’t have to think about it, it’s carried out through a magical process. This is good, considering that many cloud geeks, like myself, don’t seem to have an active part of our brains that deal with cost issues. This protects us…well…from us. Some of these things include better forecasting, a single pane of glass for multicloud, row-level access controls, cost tracking and forecasting, through the use of historical cost data, in-depth optimization recommendations, and automated remediation.