“We were essentially able to reduce the cost of that cluster by about 75%. On AWS, DevZero demonstrated they could achieve significantly higher savings than we initially thought possible.”
Mihir Nair
Head of Architecture, Databahn
Cut AWS EKS costs without
restarting your workloads
DevZero is a Kubernetes cost optimization platform that automatically rightsizes CPU and memory requests and provisions optimal EC2 instance types for EKS (Elastic Kubernetes Service) clusters, using CRIU checkpoint-restore, no pod disruption, no dropped connections.
Companies who slashed their Kubernetes
spend using DevZero
Your cluster is running, Your bill is too high
Most Kubernetes clusters run at 15–30% average utilization. Teams over-provision out of caution, pay for idle capacity, and have no clear view of what's costing money.
Requests are set once and never revisited
CPU and memory requests are set at deploy time and never touched again. Most pods run at a fraction of what they request. You pay for the full allocation regardless.
CPU allocated vs actual usage
frontend360m↔ 82% wasted
api-gateway280m↔ 72% wasted
worker1400m↔ 32% wasted
db-proxy100m↔ 84% wasted
batch200m↔ 90% wasted
0m500m1000m1500m2000m
Allocated (paid for) Actually used
Node group, m5.4xlarge × 9 (fixed, non-adaptive)
ip-10-0-1-12
m5.4xlarge · us-east-1a
19% CPU used
ip-10-0-1-47
m5.4xlarge · us-east-1a
24% CPU used
ip-10-0-1-47
m5.4xlarge · us-east-1b
11% CPU used
ip-10-0-1-203
m5.4xlarge · us-east-1a
31% CPU used
ip-10-0-1-47
m5.4xlarge · us-east-1a
24% CPU used
ip-10-0-1-47
m5.4xlarge · us-east-1b
11% CPU used
ip-10-0-4-19
·
8% CPU used
ip-10-0-4-77
·
17% CPU used
ip-10-0-5-130
·
26% CPU used
Node groups don't adapt to real workloads
Node groups are sized for peak load and stay that way. Underutilized nodes keep running with no visibility into wasted capacity and no automatic way to consolidate them.
Cross-AZ data transfer costs are invisible
AWS charges $0.01 per GB for traffic that crosses availability zones. On a large cluster, unoptimized pod placement adds thousands to your monthly bill. AWS Cost Explorer attributes the charge to EC2 totals, not to the workload that generated it.
Cross-AZ network traffic, no workload attribution
us-east-1a
frontend ×4
api-gateway ×2
us-east-1a
us-east-1a
us-east-1a
?
us-east-1b
worker ×6
postgres ×2
batch ×3
queue-consumer ×4
us-east-1a
?
us-east-1c
grpc-service ×4
kafka ×2
ml-inference ×3
event-sink ×2
log-forwarder ×1
untracked cost
One platform that actually reduces your EKS bill
DevZero deploys lightweight operators into your EKS cluster. They observe usage, generate recommendations, and apply changes automatically when you're ready. No agent rewrites. No refactoring.
Workload rightsizing, write operator
Before
frontend-pod-7f8b2
cpu: 2000m
mem: 4Gi
CPU used18%
Mem used21%
CRIU checkpoint
memory snapshot
apply + restore in-place
After
frontend-pod-7f8b2
cpu: 360m
mem: 1Gi
✓ 0 restarts
✓ connections live
Savings across 24 rightsized workloads
CPU freed: 31,440m · Mem freed: 74Gi · Applied without restart
−$8,400/mo
Rightsize requests to real usage, live
The write operator applies new CPU and memory limits in place using CRIU checkpoint-restore. Pods are never restarted, connections stay open, and you only approve changes you trust.
Node autoscaling, node operator (dzKarp)
Compute-heavy
try-restore: "true"
Memory-heavy
postgres · redis · kafka
General-purpose
workers · monitoring
dzKarp
policy-driven EC2 provisioning
c7i.4xlarge
compute-optimized
r6i.2xlarge
memory-optimized
m7i.xlarge
general-purpose
9 static nodes consolidated to 4 right-sized instances · Spot interruptions handled via SQS queue
Provision the right EC2 instance, automatically
dzKarp consolidates idle nodes and provisions cost-optimal instance types per workload class. Spot interruptions are handled via an SQS queue, managed by policy, not YAML.
Network cost visibility, network operator (eBPF)
| Workload | Type | Volume | Cost/mo | Zone |
|---|---|---|---|---|
| frontend | Cross-AZ | 847 GB | $8.47 | 1a → 1b |
| postgres | Cross-AZ | 1,204 GB | $12.04 | 1b → 1c |
| worker | Cross-AZ | 392 GB | $3.92 | 1a → 1c |
| api-gateway | Internet egress | 118 GB | $10.62 | → internet |
| Total attributed | 2,561 GB | $35.05/mo | ||
Attributed per workload via eBPF DNS tracing
See cross-AZ network cost per workload
An eBPF DaemonSet traces pod-level flows and attributes every cross-AZ and egress charge to the workload that generated it. It's the visibility AWS Cost Explorer can't give you.
Before and after with DevZero on AWS EKS
See the difference DevZero makes across workload resources, node capacity, and network costs, with the same cluster, the same workloads.
Before DevZero
Avg cluster CPU utilization22%
frontend CPU request2000m (set at deploy)
Node type selectionManual, static node groups
Spot interruption handlingManual drain scripts
Cross-AZ traffic visibilityNot tracked
Monthly infra spend$47,800 / mo
After DevZero
Avg cluster CPU utilization71% (rightsized)
frontend CPU request360m (live migration)
Node type selectiondzKarp (auto EC2 provisioning)
Spot interruption handlingSQS queue, automatic draining
Cross-AZ traffic visibilityPer-workload, per-AZ breakdown
Monthly infra spend$33,520 / mo (−30%)
Getting started in three steps
bash · prod-eks
# Install Read Operator · read-only, open-source$ helm upgrade --install zxporter \oci://...devzeroinc/zxporter \--namespace devzero-zxporter \--set config.clusterToken=<token>✓ namespace devzero-zxporter created✓ 3/3 pods running✓ metrics streaming to DevZeroread-only · no cluster changes made
devzero.io / optimization / prod-eks
24 workloads scanned−$14,280 potential
WorkloadNowRec.
frontend2000m360m
postgres4Gi1.2Gi
api-gateway1000m280m
First scan complete · No changes applied yet
Read-only · Open-source at github.com/devzero-inc/zxporter · No cluster changes until you enable automation
Deploy the open-source, read-only zxporter via Helm. See cost recommendations in your DevZero dashboard with no cluster changes needed.
bash · prod-eks
# 2a: CloudFormation · IAM + SQS queue$ aws cloudformation deploy \--stack-name "Karpenter-${CLUSTER}" \--capabilities CAPABILITY_NAMED_IAM✓ KarpenterControllerRole created✓ KarpenterNodeRole created✓ SQS: Karpenter-prod-eks ready# 2b: Install dzKarp via Helm
kubectl get nodepools · devzero-system
Kubernetes objects created
NodePooldevzero-defaultReady
EC2NodeClassdevzero-classReady
EC2 nodes provisioned
c7i.4xlargecompute-optimized×3
r6i.2xlargememory-optimized×1
SQS queue active · Spot interruption notices handled automatically · Implicit PDB enforced on all workloads
Run the CloudFormation template to create IAM roles and a Spot interruption SQS queue. Install dzKarp via Helm from the public ECR registry.
Workload policy
Scope
namespace = production
Workload types
Deployment · StatefulSet · DaemonSet
Migration type
Live (CRIU)Rolling restart
Aggressiveness
devzero.io / optimization / applying
WorkloadBeforeAfterType
frontend2000m360mlive
api-gateway1000m280mlive
worker4000m1400mrestart
Savings this month−$8,400
Automation enabled · Changes approved via policy · Savings begin immediately
Review recommendations, create workload and node policies in the DevZero dashboard, and scale down static node groups to hand off to dzKarp.
Everything AWS EKS teams need on one platform
Workload rightsizing without downtime
CRIU applies new CPU and memory limits without restarting the pod. Connections stay open. Each recommendation is labelled live or restart before you approve.
View write operator →Intelligent node autoscaling on EKS
dzKarp provisions the right EC2 instance type per workload, consolidates idle nodes, and handles Spot interruptions via SQS. Managed by policy, not YAML.
AWS EKS setup guide →Cross-AZ network cost visibility
An eBPF DaemonSet tracks pod-level flows and maps each one to AWS pricing. Covers cross-AZ, internet egress, PrivateLink, and EKS control plane traffic. No CoreDNS changes.
Network operator docs →Security scanning and compliance
Scans container images for CVEs, checks clusters against CIS Kubernetes Benchmark and NSA guidelines, and flags RBAC misconfigurations. All results in one dashboard.
Security operator →Terraform and Pulumi providers
Manage clusters and policies as code using the official Terraform provider or Pulumi SDKs for TypeScript and Python.
IaC providers →Cost-aware pod scheduling
dz-scheduler places pods on the cheapest viable nodes based on real-time Spot pricing, cross-AZ transfer cost, and CRIU compatibility.
Scheduler docs →What our customers say
Frequently asked questions
Technical questions from platform engineers who've evaluated DevZero.
Stop paying for CPU you're not usingStart rightsizing today
Connect your cluster in under 30 minutes. No code changes. No pod restarts. First saving visible within 24 hours.
Get started free