Cutting cloud waste at scale: Akamai saves 70% using AI agents orchestrated by kubernetes

Especially in this cemetery of a generative AI, cloud expenses are always at a high level. However, it is simply calculated more because enterprises are more calculated – they do not use them effectively. In fact, this year, enterprises are expected to waste $ 44.5 billion in the spending of unnecessary clouds.

This is an intensified problem Akamai Technologies: The company has a large and complex cloud infrastructure in numerous clouds not to mention numerous serious security requirements.

For Solve this, cyber work and content delivery provider turned into a KubertNetes automation platform HaveAI agents help optimize the value, security and speed in the cloud environment.

As a result, the platform, depending on the workload, helped Akamai from 40% to 70% of cloud expenses.

“We need a continuous way to optimize our infrastructure and save our cloud costs without self-sacrifice.” “We are processing working work. The delay is not an option. If we cannot respond to a security attack in real time, we failed.”

Monitoring, analysis and moving specialized agents

KeudetNetes manages the infrastructure that manages applications, facilitates them, scaling, scaling and managing brainy and microservices architecture.

Cast AI integrated customers and scales the cliques of workload, and integrated into the Kevornetes ecosystem to help manage the best infrastructure and manage the calculators and manages CEO Laurent Gil. Its main platform automation of application performance performance performance, which is constantly monitoring and operating and moving, which is constantly monitoring and moving and moving to increase the performance, security, efficiency and costs of the application (APA). Companies only reports to AWS, Microsoft, Google or others.

APA, a number of machine learning (ML) models are equipped with historical data and designed examples, and models developed by observation collection and heuristics. In several clouds combined with infrastructure-code (IAC) tools, it makes it a completely automated platform.

The clay explained that APA is a starting point in the interet; As he calls, the observation power is “not the goal.” Cast AI also supports increased setting, so customers do not need to fall and change; They can integrate existing tools and workflows. In addition, nothing leaves the customer infrastructure; All reviews and actions occur in groups of special KubertNetes that provide more security and control.

Clay also stressed the importance of human centers. “Automation completes human decision,” he said and lead the average work flows to APA.

Unique problems of ACLISH

Shavit Akamaı’nın explained that it was great and complicated cloud infrastructure Significant service agreements (the world’s most demanding customers and industries “and cyber-cienton services in accordance with SLA (SLA) and performance requirements.

He noted that for some services they consume, the largest customers for the sellers, the “major engineering and reenrying” for sellers, the hyperscaleri said they did to support their needs.

In addition, Akamai serves customers of various sizes and industries, including large financial institutions and credit card companies. The company’s services are directly related to customers with security posture.

As a result, Akamai was needed to balance it with the cost of all this complexity. Shavit said that real life attacks on customers can hold 100x or 1000x, the specific components of the infrastructure. But “Pre-cloud ability to calculate our ability in advance is not only financially possible,” he said.

The team was considered optimizing on the code, but the unique complexity of their work model was the main infrastructure.

Automatically optimizes all Kubernete infrastructure

Aycamai was a Kubernetes automation platform that you really need to optimize all its costs The main infrastructure In a few clouds in real time, explained the shavel and constantly applied applications on the basis of changing demand. But all this was to be done without sacrificing the application performance.

Before applying the casting, Shavit noted that Akamai’s DevoPs team manually regulates all Kubertnetes working loads a month. Given the scale and complexity of the infrastructure, it costs difficult and expensive. Only by analyzing workloads in sporadically, they missed any real time optimization potential.

“Now hundreds of collapsing agents are the same tuning, except, each day is every second,” he said.

Core APA’s use Aycamai use automation, bin packaging (minimizing the number of containers), automatic selection of most economical calculation instances, workload, spot instance automation and expenses analytical capabilities.

“We’ve got an idea for an analyst with an analyst with something we have ever seen before,” he said. “Optimization has started automatically after active agents have been placed and the deposits began to enter.”

Spot instances – where the enterprise may include unused cloud capacity – clearly in the business sense, but the complex workloads of the Aklaman appeared to be complicated due to Apache Spark. It was needed to put the overengineer workload or more processing hands appeared to be in the opposite of the material.

With Cast AI, they were able to use the engineering group or transactions with “zero investment” with “zero investment”. The value of spot instances was “super clear”; Needed to find the right tool to be able to use them. Shavit noted that this was one of the reasons why they were moving forward.

While saving 2x or 3x in cloud bill, Shavit pointed out that automation was “priceless” without interference. This caused “mass” time savings.

Before performing Cast AI, his team is constantly moving in circles and keys to ensure that the production environment and customers need to invest. ”

“The greatest benefit, the greatest benefit we need to be managed by our infrastructure,” he said. “Cast’s team agents now make it for us. This has released our team to direct our team to the most important ones: to provide our customers faster.”

Editor’s note: This month Turn a vbGoogle Cloud Cto, Grannis and Highmark Health SVP and General Analytical Officer Richard Clarke, Multi-adjustable environment will discuss many Model AI systems in a healthcare area and real world problems. Register today.