Proven Ways to Use AI to Cut Cloud Costs
AI can reduce cloud costs in two key ways.
- By David Drai
- March 7, 2022
Organizations are learning how they can reap several benefits from deploying artificial intelligence (AI) tools. For example, they can complete processes and tasks faster through automation, boost productivity and operational efficiency, gain a better understanding of customers to improve their experiences, and make faster and more informed business decisions.
A lesser-known benefit of AI is that it can help reduce cloud costs, something organizations need to be concerned about as they move more heavily into the cloud for virtually every aspect of IT.
The proportion of IT spending shifting to cloud will accelerate in the aftermath of the COVID-19 pandemic, Gartner says, with cloud projected to make up 14 percent of the total global enterprise IT spending market in 2024, up from 9 percent in 2020.
AI can reduce cloud costs in two key ways: by automatically discovering cost spikes and glitches in cloud computing; and by forecasting future cloud costs to aid an organization’s planning.
Finding Spikes and Glitches
Cost spikes often occur when someone makes a mistake or uses a cloud resource at a much higher rate than required. It’s like returning home from a weekend trip only to find you left the air conditioner on full blast. Both waste resources and can wreak havoc on your monthly bill. Organizations need to discover wasteful cloud spending quickly or risk significant fluctuations in their cloud costs.
The cloud is, basically, a utility and it behaves as such. Many components of IT can expend cloud computing resources and run up costs in a hurry. These include servers, network traffic, database queries, and cloud services.
Organizations have become highly dependent on cloud computing, and the work-from-home model resulting from the pandemic has increased this dependency significantly over the past year, increasing the potential for such glitches. In fact, Gartner has reported that companies that are unaware of the mistakes made in their cloud adoption will overspend by 20 to 50 percent.
One example of such a glitch happened at my company, Anodot. A test we ran was supposed to spin up only a few machines but spun up 100 machines instead. This led to additional cloud costs of more than $10,000 for this one incident. This happens more often than you’d think at organizations, no matter how many people are monitoring their cloud services.
Queries to cloud-based data services are another source of rising costs. With services such as Google Cloud Platform’s (GCP) BigQuery database, some of the charges are based on the queries requested. A typical monthly GCP bill might be $115,000 per month, but frequent BigQuery queries can drive the cost up by thousands of dollars.
AI solutions can help organizations avoid these extra costs. For example, a company conducting frequent and unnecessary queries to cloud data services might not notice the problem for days or weeks. Algorithms work independently to find anomalies in just a few hours so teams can remediate quickly.
Automating the monitoring and detection of cloud usage anomalies also help alleviate wasted work-hours. Many companies still use labor-intensive processes to analyze usage, such as checking dashboards manually. The larger the company (and thus the team), the greater the waste.
Finally, there’s the issue of accuracy. AI-based solutions can monitor multiple services and do so in greater detail, learning the behavioral patterns to better distinguish what’s normal usage for a given time of day or week and what’s an anomaly.
Forecasting Future Costs
Organizations can also leverage AI to forecast future costs of different cloud services. AI-based solutions can help more effectively and proactively manage an enterprise’s overall spending on cloud services.
AI-powered tools can analyze whatever factors are driving costs and suggest solutions. AI and forecasting can also help organizations optimize cloud usage by types of services. With the on-demand capabilities of the cloud, companies can use AI to predict demand and then make reservations for when they want to use certain applications or storage resources. In this way they can achieve greater optimization, which leads to reduced costs. This is changing the way software works and the way organizations use services.
An AI-based forecast shows the anticipated overall spending for cloud services, but takes it a step further by breaking costs down by service to show how much each will consume in the coming month. AI algorithms can factor in historical data and fluctuations in current usage to more accurately anticipate future cloud usage than teams can manually.
Furthermore, with AI-based forecasting, enterprises can monitor usage to catch mistakes as early as possible. They can determine when to use compressed rather than uncompressed files, which could save storage costs by a significant amount.
Putting AI to Work
Manual cloud cost optimization is a time-consuming and complex process at a high engineer time cost. AI-based monitoring and forecasting enables teams to take a proactive approach to all things cloud, yet not many IT leaders are leveraging it for this purpose.
Any type of software-as-a-service (SaaS) application that is billed by consumption is a natural candidate for AI. That even includes applications such as Slack that have dynamic billing.
Organizations can be reluctant to use AI to help with cloud cost cutting because they either aren’t aware AI can be used this way or because past AI project failures have made them skeptical. However, with a new crop of proven AI solutions and the increasing dependence on cloud services, that hesitancy needs to change. Whether it’s by discovering spikes and glitches or forecasting costs, AI can help you reduce cloud costs and gain the most value from your cloud services.
David Drai is the CEO and cofounder of Anodot, a tool for automated anomaly detection and real time analytics. In his career, Drai has served as the CTO of Gett Taxi, and Contendo CTO and cofounder, which was sold to Akamai in 2012. You can reach the author at firstname.lastname@example.org.