CPU/Mem reactive · K8s default
Pod CPU/Memory
↓
threshold >70%
↓
Replica ±1 · ~30s
- ✕Usage ≠ performance
- ✕Reactive only
- ✕Manual tuning
Run Kubernetes
with top performance at minimal cost
Wave is an integrated Kubernetes solution for scaling, sizing, traffic management, diagnostics, and node management.
It addresses operational reliability challenges that Kubernetes platforms such as Amazon EKS and OpenShift do not cover.
Wave
Wave is a Kubernetes operations platform that automates autoscaling, resource sizing, incident diagnosis, and node provisioning in one integrated workflow.
THE PROBLEMThe moment you assume one-time setup is enough,
The moment you assume one-time setup is enough,
costs start to creep up.
Problems in K8s environments pile up quietly. Unnoticed waste, unannounced outages, unfair traffic handling that erodes trust — the damage lands squarely on the business.
Hidden infrastructure costs
65% of containers use less than half of their requested resources. You pay every month for capacity you never use.
Scaling that breaks at every peak
Even when HPA reports healthy, the service can stall — the same failure that repeats at every flash sale.
A traffic spike is a revenue loss
Critical traffic is handled at the same priority as batch jobs. You can lose a million dollars in an hour.
HPA sees CPU only,
Autopilot
learns the workload
Autopilot scales pods with a performance model that learns the service's real behavior.
It decides in about 10 seconds, roughly 2× faster than HPA, with stability guardrails built in.
Autopilot
Scheduler
Min/Max
Workload-aware · ML model
latency · throughput
↓
ML auto-decision
↓
accurate · ~10s
- ✓Performance model
- ✓~2× faster than HPA
- ✓No threshold tuning
Pod requests, exactly right — nodes cut in half
Wave Sizing learns each workload's real demand and consolidates 10 nodes into 4.
It keeps headroom while cutting infrastructure cost by about 60%.
Smart Sizing
Right Sizing
Without right-sizing
With right-sizing
Free
Free
Free
Free
Free
Free
oversizedright-sizedfreed
Fast Karpenter,
node cost down about 70%
Node Warmup removes cold-start delays, while Spot Workload Placement keeps stateful and critical pods off interruptible nodes. Reduce node costs by about 70%.
Dashboard
Node Warmup
Spot Placement
K8S SCHEDULING PIPELINE→→→Wave
Pod Created
Pending
+30~90s✗
Karpenter
provision
Scheduled
ready✓
Node Warmup
Pre-provisions nodes ahead of demand — Pending step eliminated
RESULT
~70%↓
node costDashboards show metrics,
Diagnosis points to the cause
Five ML diagnostics (resource forecast · memory leak · CPU saturation · scheduling delay · idle node) surface risk before the alarm ever fires.
Forecast
Memory Leak
CPU Analysis
Sched Delay
Idle Node
Grafana shows metrics. People still guess what went wrong.
ML-driven forecasting, detection, analysis.
Catch leaks weeks before OOMKill. ✓
Keep checkout running,
scale the rest down gradually
Four-tier priority-based traffic shaping on top of Istio handles critical traffic first and progressively throttles the rest as load rises. A spike in one service is contained before it becomes a broader outage.
Wave Flow
Without Wave Flow · Istio default→
→
Istio
no priority
✗✗✗✗✗
Checkout ~30%
DROP ✗
With Wave Flow · priority shaping→
→
CRITICAL→ pass
STANDARD→ queue
BACKGROUND→ defer
CRITICAL PROTECTED
Checkout 100% ✓
PV operations, without tickets
It forecasts capacity, expands volumes before pods stall, and reclaims unused PVs. Zero manual operations.
PV Auto-Expansion
Capacity Forecast
PV Cleanup
Unused PV Detection
PERSISTENT VOLUMES · 4
USED
Wave Action
data-postgres-0
87%
+ AUTO-EXPAND ✓
redis-cache
22%
FORECAST · 14d
logs-archive
0%
→ CLEANUP (31d idle)
backup-vol
45%
healthy
zero ops tickets
Works with your
existing stack
A programmable API for CI/CD and Terraform, plus transparent alerts delivered through Slack and webhooks.
Programmable API
Alerts
POST /v1/clusters "autopilot": true "min": 2 "max": 20 200 OK ✓
→
Wave
→Slack alerts
→Webhook
→Terraform · CI/CD
Frequently Asked Questions
Common questions before getting started. Reach out anytime to learn more.
Wave is a Kubernetes operations platform that integrates and manages Kubernetes scaling, resource sizing, incident diagnosis, traffic control, node provisioning, and more. Wave is not a simple autoscaling tool; through features such as Autoscale, Sizing, Diagnosis, Flow, Karpenter, and GPU, it manages the performance, cost, and stability issues that arise in Kubernetes operations.
써 보는게 가장 빠릅니다.
서비스 환경 변화 없이 원하는 방식에 따라 빠르게 테스트 및 적용 가능합니다
(Server Side 에이전트의 경우 별도 적용 요청 필요)