Wave Autoscale 2.6.0 Release
Author: Hwansoo Kim (opens in a new tab)
Date: June 24, 2025
Wave Autoscale 2.6.0 expands our Kubernetes optimization engine with deeper metric insights, smarter traffic control, and more efficient resource modeling. This release empowers teams to operate infrastructure more autonomously—especially in high-scale environments with Istio and multi-container workloads.
📡 Istio Ingress Metrics Collection
Automatically collects requests and latency metrics directly from Istio Ingress Gateways.
- Enables traffic-aware scaling decisions by connecting metrics to Kubernetes Deployments automatically.
- No manual instrumentation or custom labeling required.
New Feature: 🌊 Wave Flow
Wave Flow is a traffic shaping orchestration layer within Wave Autoscale. It enables users to configure, manage, and automate fine-grained traffic control across Ingress Gateways and Service Mesh proxies, powered by real-time infrastructure metrics and Proxy-WASM module (Wave Shaper).
✅ Purpose:
- Protect critical services under load.
- Optimize system resilience.
- Automate adaptive load shedding.
✅ Wave Flow tightly integrates:
- Kubernetes-native workloads (Ingress, Service Mesh, Waypoints)
- Wave Autoscale's real-time metrics (CPU, memory, I/O)
- Proxy-WASM extensions (via Wave Shaper)
Wave Shaper classifies every incoming request into one of four priority classes:
| Priority Class | Description | Priority |
|---|---|---|
| CRITICAL | Essential business-critical operations (e.g., checkout, login, payment) | 🥇 Highest |
| IMPORTANT | High-value operations that impact business or key user experience | 🥈 High |
| MODERATE | Useful features that improve UX or performance but are not essential | 🥉 Medium |
| BULK | System background tasks (e.g., batch analytics uploads) | 🥉 Lowest |
⚙️ Direct Kubelet Metrics Access
Metrics are now retrieved directly from Kubelet, bypassing the API server.
- Faster metric delivery.
- Reduced API server load and improved resilience.
🧹 Container Metrics Exclusion
Exclude specific containers (e.g., sidecars, init containers) from autoscaling logic and ML training.
- Improves signal clarity.
- Supports annotation-based exclusion.
🤖 Enhanced Autopilot ML Performance
Autopilot's ML models are now more effective in low CPU and memory usage environments.
- Improved prediction under light load.
🛠 Minor Fixes and Improvements
- Stability improvements in WA Metrics Agent.
- Dashboard responsiveness and UX polish.
- Internal model tuning and metadata handling fixes.
Wave Autoscale 2.6.0 continues our mission to automate infrastructure operations with precision and resilience—offloading manual effort and enabling smarter decisions at every scale.