Join the event entrusted by business leaders for almost two decades. VB Transf transform gathers the people who build a real strategy of AI. Get more information
In particular, in this age of generative, the costs of the cloud are always high. But this is not only because companies use more calculation, they do not use it efficiently. In fact, only this year, companies are expected to waste $ 44.5 billion in unnecessary cloud costs.
This is an amplified problem for Akamai technologies: The company has large and complex cloud infrastructure in several clouds, not to mention numerous strict security requirements.
Into Solve -Cybersecurity supplier and content delivery went to the Kubernetes Automation Platform Distribution you haveIA agents help optimize cost, safety and speed between cloud environments.
Ultimately, the platform helped Akamai reduce 40% to 70% of cloud costs, depending on the workload.
“We needed a continuous way to optimize our infrastructure and reduce the costs of our cloud without sacrificing performance,” said Dekel Shavit, lead director of cloud engineering in Akamai, in Ventubbeat. “We are the ones who process security events. Delay is not an option. If we cannot respond to a real -time security attack we have failed.”
Specialized agents who control, analyze and act
Kubernetes manages the infrastructure that manages applications, facilitating the deployment, climbing and managing -especially in native of the cloud and architectures of microserveis.
Cast Ai has joined the Kubernetes ecosystem to help clients climb their clusters and workloads, select the best infrastructure and manage the computing life cycles, explained the founder and CEO Laurent Gil. Its basic platform is Application Performance Automation (APA), which operates through a team of specialized agents who control, analyze and take steps continuously to improve the performance of the application, safety, efficiency and cost. Companies only provide the calculation they need of AWS, Microsoft, Google or others.
APA is fed by various automatic learning models (ML) with reinforcement learning (RL) based on historical data and learned patterns, improved by a pile of obsability and heuristics. It is accompanied by infrastructure tools as a code (IAC) in various clouds, which makes it a completely automated platform.
Gil explained that APA was built on the principle that observability is just a starting point; As he called it, obsability is “the foundation, not the goal.” AI distribution also admits incremental adoption, so customers do not have to leave and replace; They can integrate into existing tools and flows of work. In addition, nothing never leaves customer infrastructure; All analyzes and actions occur within their clusters dedicated to Kubernetes, providing more security and control.
Gil also emphasized the importance of human centering. “Automation complements human decision -making,” he said, with APA, maintaining human work flows in the middle.
Akamai’s unique challenges
Shavit explained that the great and akamai complex Cloud infrastructure Content delivery power (CDN) and cybersecurity services provided to “some of the most demanding clients and industries in the world” while fulfilling strict service levels (SLA) and performance requirements.
He said that for some of the services they consume, they are probably the older customers of their seller, adding that they have made “tons of engineering and reengineering” with their hyperscaler 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 the security stance of their customers.
Ultimately, Akamai needed to balance all this complexity with cost. Shavit said that real -life attacks on customers could lead the capacity of 100x or 1,000x in specific components of their infrastructure. But “climbing our capacity for clouds for 1,000 times ahead is not possible financially,” he said.
His team planned to optimize on the side of the code, but the inherent complexity of his business model required to focus on the main infrastructure.
Automatically optimizing all Kubernetes infrastructure
What Akamai really needed was a Kubernetes automation platform that could optimize the costs of the execution of everything Basic infrastructure In real time in various clouds, Shavit and scale the applications up and down according to the constantly changing demand. But all this had to be done without sacrificing the performance of the application.
Prior to implementing the cast, Shavit said that the Akamai Devops team manually adjusted all their Kubernetes workloads rarely a month. Given the scale and complexity of its infrastructure, it was difficult and expensive. Only by analyzing the workloads sporadically, did they clearly lost any real -time optimization potential.
“Now, hundreds of distribution agents do the same tune, unless they do it every second of each day,” Shavit said.
The basic features that use Akamai are automation, automation of in -depth Kubernetes with the packing of the bins (minimizing the number of bins used), the automatic selection of the most profitable calculation instances, the rights of workload, the automation of specific instances throughout the instance of the instance and the capacities of cost analysis.
“We have been aware of the two -minute cost analytics on integration, which we never saw before,” Shavit said. “Once the active agents were deployed, the optimization started automatically and the savings began to enter.”
The timely instances: where companies can access the unused cloud capacity at discount prices, they obviously made sense of business, but they were complicated due to the complex Akamai workloads, especially Apache Spark, said Shavit. This meant that they needed workloads above or put them more in their hands, which turned out to be financially opposite.
With the distribution, they were able to use timely instances in Spark with “Zero Investment” of the engineering team or operations. The value of the specific instances was “super clear”; They just needed to find the right tool to be able to use them. This was one of the reasons why they advanced with the cast, said Shavit.
Although saving 2x or 3x on the cloud bill is excellent, Shavit said that automation without manual intervention has no “price”. It has resulted in a “massive” time saving.
Before implementing the AI distribution, his team “was constantly moving by knobs and switches” to ensure that their production environments and their customers were up to date on the service they needed to invest.
“The biggest benefit has been the fact that we no longer need to manage our infrastructure,” Shavit said. “The team of the Cast agents now does it for us. This has released our team to focus on what matters most: freeing functions faster on our customers.”
Publisher Note: In the month Transform vbGoogle Cloud Cto Will Grannis and Highmark Health SVP and the head of Analytics Chief Richard Clarke will discuss the new AI stack in healthcare and the real world challenges of deploying multimodel systems in a complex regulated environment. Sign up today.
Source link
