Skip to content
Karpenter offers a lot of advantages over Cluster Autoscaler, enhancing pod placement for improved instance utilization and lower compute costs. It also helps seamlessly integrate cost-effective Spot instances.
However, Karpenter isn’t perfect, there are some challenges:
- 2-minute Spot interruptions disrupting workloads
- The complexity of determining the optimal blend of Reserved Instances, Savings Plans, and Spot
- Infrequent reassessment of Kubernetes configurations leading to inefficiency
In this blog, we’ll talk about how nOps makes Karpenter even better — and how easy it is to add nOps to your existing Karpenter clusters.
The Karpenter Advantage
EKS Cluster Scaling
Native Integration with AWS APIs
Karpenter (and nOps) integrates natively with a broad range of AWS APIs in order to perform auto-scaling actions in EKS clusters. Out of the box, it already supports a diverse range of instance selection options, which are configurable via NodePools and NodeClasses. These include, but are not limited to instance capacity types (On-Demand, Spot), family, size, generation, architecture and OS.
Node Rightsizing
Karpenter addresses pre-defined node group constraints associated with traditional tools like Cluster Autoscaler, enabling more fine-grained control over resource utilization. This is accomplished via NodePools, which allow users to specify a wide variety of constraints on nodes, such as instance groups, families and/or sizes, availability zones, architectures, and capacity types – allowing Karpenter to make optimal decisions on what instance to start or terminate next.
Native Support for Spot Termination
Karpenter automatically handles Spot interruption notifications by consuming events from an Amazon SQS queue, populated through Amazon EventBridge. When Karpenter receives a Spot interruption notification, it gracefully drains the interrupted node of any running pods while also provisioning a new node onto which those pods can be rescheduled. As long as this can be completed within 2 minutes, operations continue smoothly without significant downtime or performance impact.
Why Karpenter and nOps are better together
While Karpenter offers significant advantages over Cluster Autoscaler, it does not inherently possess certain cost optimization capabilities, such as:
- Awareness of your EC2 Reserved Instances and Savings Plan commitments
- Advanced analysis of EC2 Spot pricing data
- Enhanced EC2 Spot Instance termination prediction
By adding these enhancements, nOps Compute Copilot empowers you to run many more workloads safely on Spot, for greater savings. It continually tunes your Karpenter configurations to minimize interruptions, maximize savings, and maximize your commitment utilization.
nOps manages $1.5+ billion in cloud spend. By analyzing massive amounts of proprietary Spot market and historical data with Machine Learning, nOps can predict Spot terminations 60 minutes in advance. It continually monitors your usage, Spot pricing and Spot availability to proactively move workloads onto reliable and cost-effective instance types and drastically reduce Spot interruptions.
Cluster Performance Dashboard
Navigating the AWS console to gather comprehensive information about your resources often requires clicks and calculations for potentially hundreds or thousands of clusters.
nOps adds an intuitive UI to Karpenter, making it easy to visualize how your workloads are being optimized in real time. All of the essential information is consolidated into one unified dashboard, so you don’t have to hunt down individual resources. All in one place, track how much you’re saving, see your real hourly resource cost, and verify the state of each EKS cluster and the instances it’s running on.
Intelligent Instance Selection Based on Workload Requirements
Not sure what instance families or sizes will suit the specific requirements of your workloads? With nOps Copilot, engineers no longer have to sift through the vast sea of AWS instance types.
Just specify your workload requirements once, and we will automatically narrow down the right instance types so you never have to worry about compatibility.
Guided Karpenter Configuration and Continuous Tuning
Karpenter is often tuned as a one-and-done exercise — but your ecosystem is dynamic. Optimal configuration of Karpenter is strongly interlinked with the state of the compute in the cluster, meaning that Karpenter configurations need to be continuously revisited to ensure that they are optimal.
Automation allows Copilot to tune Karpenter much more frequently than a human maintainer would — translating to better results and many hours of work saved.
The Ultimate Guide to Karpenter
Real-time instance reconsideration with RI and Savings Plan Awareness
Your workloads and the Spot market are always changing, meaning the best instances to run on are too. Staying ahead of these constant developments is difficult if not impossible to do manually.
nOps enhances Karpenter by adding awareness of the Spot Market and your existing AWS purchase commitments. As clusters scale, Spot availability changes and commitments become overutilized or underutilized, nOps continually monitors your workloads to ensure you are always on the most cost-optimized and stable instances.
Automated Provisioning with RI, Savings Plan, and Spot Awareness
Ensuring you use all of your AWS purchase commitments can be a challenge, particularly as they apply hourly on a use-it-lose-it basis. It’s complex and difficult to manually determine the perfect blend of RI, SP and Spot to use for optimal price and stability.
Copilot ensures you’re always on the right balance, automatically updating your provisioning on the fly if not.
Automated Commitment Management
Forecasting commitments is an art. Predicting how much and what kind of RI and SP to buy can be complex and risky.
Copilot fully manages commitments on your behalf so you don’t have to manually review and adjust purchase decisions, saving time and ensuring commitments never go unused. Plus, a buy-back guarantee eliminates your risk of over-committing — so you get more savings AND more flexibility.
Intelligent Spot Management that Maximizes Stability and Scalability
Understanding Spot market pricing and termination risks and choosing instances to balance cost, performance, and availability demands a deep understanding and expertise.
nOps makes it orders of magnitude easier and more reliable to use Spot than with Karpenter alone. It analyzes your workloads and the Spot market to generate a broad and diverse list of suitable instance families, such that there are always cheap and reliable instances available. It automatically and continuously moves your workloads onto optimal instances while gracefully draining nodes in the process, drastically reducing the amount of involuntary Spot interruptions. In fact, nOps offers the same SLAs as AWS On-Demand.
Karpenter + nOps are the perfect combination
nOps Compute Copilot built on Karpenter is designed to make it simple for users to maintain stability and optimize resources efficiently. To maximize EKS cost and performance, automation is key. nOps helps engineering teams to more easily and effectively leverage the power of Karpenter and Spot for cost savings and reliability, freeing them to focus on building and innovating.
nOps was recently ranked #1 in G2’s cloud cost management category. Book a demo to find out how to save in just 10 minutes.
