Skip to content
Network IntelligenceEarly access

What your cross-zone and internet traffic costs, per workload.

An eBPF agent on each node counts bytes between pods, zones and the internet, and Kubeadapt prices the traffic per workload. No sidecars, no pod restarts.

One pod per node · Linux 5.10+ · AWS pricing today
Example · prod-us-east-1 · last 30d
$2,620 / mo · 215 TB
us-east-1a
42 TB
$362
18.1 TB
$184
9.2 TB
$360
4.0 TB
us-east-1b
$220
11 TB
48 TB
$486
24.3 TB
$270
3.0 TB
us-east-1c
$124
6.2 TB
$264
13.2 TB
32 TB
$350
3.9 TB
Same-AZ · freeInternet egressLargest flow
Source AZ → Destination AZ · last 30d
Largest cross-zone flow
analytics-worker us-east-1b → us-east-1c
$486/mo· 24.3 TB

CEL · query language

Query your network
like a database.

Filter flows with CEL, the expression language Kubernetes uses in ValidatingAdmissionPolicy and CRD validation rules.

Filter by

NamespaceWorkloadPodServiceZoneNodePortProtocolDestination domainCost typeBytesCost

Cross-zone and internet traffic

Traffic between zones and to the internet is billed per GB.

On AWS, cross-zone traffic costs $0.01/GB in each direction, and internet egress up to $0.09/GB. Traffic inside a zone is free. Below: one example cluster’s traffic for 30 days, by source and destination zone.

Cross-zone + egress, 30 days
$2,620
215 TB · example · prod-us-east-1
Source AZ → Destination AZ · 30 days
Same-AZCross-AZInternet egress
us-east-1aus-east-1bus-east-1cus-east-1aus-east-1bus-east-1cInternetSOURCEDESTINATION$362/mo$360/mo$486/mo$270/mo$350/mo
Same-AZFree
$0
/mo
122 TB $0.00/GB
Cross-AZBoth sides billed
$1,640
/mo
82 TB $0.01/GB each way
Internet egressPer GB out
$980
/mo
10.9 TB up to $0.09/GB

Top cross-zone talkers

Which workloads send traffic across zones.

Workloads ranked by cross-zone cost, with the zone most of their traffic goes to. What to change, if anything, is your call.

Cross-zone cost, top 4
$1,332/mo
#1
analytics-worker
runs in us-east-1b · 24.3 TB
97% to us-east-1c
$486
/mo
#2
checkout-api
runs in us-east-1a · 18.1 TB
88% to us-east-1b
$362
/mo
#3
inventory-svc
runs in us-east-1c · 13.2 TB
82% to us-east-1b
$264
/mo
#4
ml-inference
runs in us-east-1b · 11 TB
93% to us-east-1a
$220
/mo

Example data · last 30 days · AWS transfer rates

Architecture

One agent per node. No sidecars.

A service mesh needs a proxy in every pod. Kubeadapt runs one eBPF agent per node and counts bytes per connection, so you get cost per flow without touching your pods. It records addresses, ports, byte counts, DNS names, TLS SNI and HTTP method, route and status. It never records payload bodies.

Traditional · sidecar per pod

12 pods, 12 proxies

1 proxy / pod
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
APP
+PROXY
Each pod injected · rolling restart required12 extra containers
Kubeadapt · DaemonSet + eBPF

12 pods, 1 agent on the node

~150 MiB / node
APP
APP
APP
APP
APP
APP
APP
APP
APP
APP
APP
APP
kubeadapt-k8s-pulseeBPF programs
LINUX KERNEL · 5.10+no kernel modules
No pod injection · no pod restarts1 pod per node
Memory
~150 MiB
measured per node · 256 MiB requested
Linux kernel
5.10+
no kernel modules
Deployment
DaemonSet
1 pod per node
Runs as
root
privileged · hostNetwork · hostPID · read-only root filesystem

Requires the network agent kubeadapt-k8s-pulse, installed with the same Helm chart (--set kubeadapt-k8s-pulse.enabled=true). The DaemonSet runs privileged, with hostNetwork and hostPID, as root with a read-only root filesystem. Prices: AWS today. GCP and Azure: not yet.

Early access

Try it on your own clusters.

Early access is by request. Book a call and we will walk you through it. Cost monitoring, rightsizing and the rest of Kubeadapt are available today.

First 100 vCPU freeNo credit card