Zitec Blog

Zitec x AWS Event Highlights: Amazon Elastic Kubernetes Service as a cutting-edge, cost optimization solution for today's cloud operations

Written by Andrei Radu | Jul 4, 2023 9:38:20 AM

Having the right and most efficient strategies, technologies, and processes in place is critical for businesses to succeed and optimize their costs. In light of this, Zitec and Amazon Web Services joined forces to organize the Challenges & Cost Optimization for Amazon EKS Workloads event on May 24th, live at Zitec HQ. 

This event served as yet another significant milestone in the longstanding partnership between Zitec and AWS, demonstrating over 15 years of expertise in AWS practices and the successful deployment of numerous AWS-based solutions.

Two highly informative presentations held by Zitec and AWS experts showcased how to harness the power of Amazon’s Elastic Kubernetes Service (EKS) to increase business scalability, optimize their cloud spending, and improve resource allocation in day-to-day operations.

For those of you who missed out on the event, we’ve gathered below some of the highlights, by elucidating two of the most poignant aspects of EKS: optimizing computing capacity and costs.  

How to optimize computing capacity through planning

Daniel Chiricheș, DevOps Engineering Manager, Zitec

The event kicked off with a presentation held by Daniel Chiricheș, DevOps Engineering Manager at Zitec, who highlighted several key practices that businesses and teams could implement to optimize costs when deploying and managing an AWS infrastructure:

  • define requirements and goals, and pick the right-size instance types: understanding the expected workload, application characteristics, performance needs, and scalability requirements, and choosing the appropriate Amazon EC2 instance type is a must;
  • monitor and analyze existing workloads' resource usage patterns, such as CPU, memory, and storage;
  • estimate future requirements and continuously monitor for changes: based on expected growth, estimating future resource requirements and adapting to changes to the plan should always be kept in mind;
  • use Autoscaling: using EKS Horizontal Pod Autoscaler capabilities (HPA) to automatically adjust the size of clusters based on demand. 

Carefully considering these suggestions during the deployment of a Kubernetes-based infrastructure and its day-to-day operations optimizes resource allocation and lowers costs. 

Where can you optimize costs with EKS best practices

Radu Dobrinescu, Senior Partner Solutions Architect CEE, Amazon Web Services

AWS Partner Solutions Architect Radu Dobrinescu delivered an engaging presentation on how EKS can significantly enhance cost optimization across three areas:

  • Purchase and savings options: Radu Dobrinescu shared valuable insights on how businesses can leverage EKS to explore different purchasing models and savings options within the AWS ecosystem. By making informed choices and taking advantage of cost-saving opportunities, organizations can effectively reduce their overall cloud expenditure while maximizing value;
  • Autoscaling mechanisms: Another vital topic covered by Radu Dobrinescu was the utilization of EKS's autoscaling capabilities. Attendees discovered how this powerful feature empowers businesses to adjust their resources based on demand dynamically. By automatically scaling up or down in response to workload fluctuations, organizations can optimize their cost efficiency while ensuring optimal performance;
  • Visualization of spending: Radu Dobrinescu emphasized the importance of visualizing spending patterns within EKS. By leveraging comprehensive monitoring and reporting tools, businesses gain valuable insights into their resource utilization and can identify opportunities for further cost optimization. Clear visualization enables informed decision-making and proactive cost management.

How to reduce costs with Amazon Elastic Kubernetes Service

One way to optimize costs is by using software containers - lightweight software packages that have all their dependencies within. Containers allow businesses to cut costs with OS licensing and increase compute utilization through their bin-packing and are poised to be ”the default choice for 75% of new customer enterprise applications by 2024”.

Costs can be further reduced by opting for AWS Managed Containers to remove the heavy lifting required to manage a network and free up part of the budget that would go towards paying container licensing fees and control plane maintenance. 

Radu Dobrinescu used a sample expense comparison in order to better portray the differences between using EKS and Self Managed K8s, using the following assumptions:

  • Small app workload across 5 clusters with 50 nodes per cluster;
  • m4.xl assumed as Control Plane Compute instance type. No savings plan applied;
  • Rough estimate not including all associated services and cost such as storage and logs;
  • Resource effort (FTE) savings not calculated. 
USING EKS SELF MANAGED K8s
AWS Control Plane cost: $-4k/yr for 5 clusters ($876/clusters/yr with $0.10 per cluster) AWS Control Plane cost: $-44k/yr for 5 clusters ($∼9k/K8s Cluster/yr for compute alone)
AWS Support Cost: EKS support cost included AWS Support Cost: Doesn’t include K8s support

How to lower computing costs within EKS

Computing costs made up over 80% of the cost within Kubernetes (K8s) clusters, according to data from the FinOps Foundation. To further reduce costs for Amazon Elastic Kubernetes Service users, Radu identified four main points that businesses and IT professionals could address to lower computing costs:

  • choosing the right purchase option for your operation needs - with EC2 spot instances being able to cut costs with up to 90%;
  • scaling and optimizing your infrastructure through the use of out-of-the-box EKS auto-scaler or Karpenter for larger Kubernetes clusters;
  • monitoring cost in real-time with granular options and with the aid of Kubecost - which Amazon EKS users can get without any additional licensing cost;
  • utilizing AWS-designed Graviton chips which are fine-tuned towards AWS cloud-based workloads and offer better price for their performance. Customers can leverage multi-architecture (ARM64 and AMD64) container images to run mixed workloads on Graviton and x86 instances;

EC2 Spot Instances are one of the main purchase options, alongside On-Demand and Reserved Instances. 

What makes Spot Instances so attractive cost-wise is the ability for users to utilize spare EC2 instances at a substantial discount - which can reach up to 90%, compared to the other options, for as long as those instances are not needed by On-demand users.

”Time-flexible” interruptible workloads are prime candidates for Spot Instances, which can leverage available EC2 instances from diverse availability zones and varying types.

Cost optimization with the help of Karpenter and Kubecost

Optimizing K8s services expenditure can be further achieved through automating scalability needs and granular cost monitoring with the help of specialized open-source software:

Radu’s presentation showcased:

  • Karpenter as as a high-performance open-source autoscaler that addresses scalability challenges such as right-sizing instances, efficient node management and built-in handling of spot interruptions;
  • Kubecost and AWS Cost Explorer, for monitoring cluster costs, with Kubecost providing a more granular view on the matter.

Karpenter’s main strength as an autoscaling tool comes from its ability to deploy and optimize nodes quickly, within seconds - its speed translates into more efficient use of EC2 instances by minimizing wasted time during scaling operations.

Meanwhile, Kubecost offers a granular approach to cost monitoring of K8s services. Kubecost allows users to see expenditures in real-time and gain insights for Kubernetes at a Namespace and Pod level and even at the organizational level - business units or teams. 

Kubecost's features come without any further costs with licensing for Amazon EKS customers and with added support from AWS.

Conclusions

The Challenges & Cost Optimization for Amazon EKS Workloads event marks another successful milestone in Zitec and AWS's long-standing collaboration. Speakers from both companies delivered presentations on the latest cost optimization solutions for users of Amazon Elastic Kubernetes Service through the employment of cutting-edge software solutions and best practices.

Stay tuned for more DevOps expertise and insight from us and our partners! If you require further information on the best EKS practices from an AWS-certified DevOps team or want to learn more about best DevOps practices, you are welcome to contact us through our DevOps Services pages.