Overview

The following tutorials explain how you can benefit from using Kedify in various use cases:

HTTP Scaling Guides

The guides in this section cover various HTTP scaling strategies and configurations, helping you choose the best approach for your application’s deployment scenario.

Basic Scaling Guides

• HTTP Scaling for In-Cluster Traffic (Kubernetes Service) - For applications that do not expose an external Ingress and rely solely on internal Kubernetes Services.
• HTTP Scaling for Ingress-Based Applications - For applications exposed using Kubernetes Ingress.
• HTTP Scaling for OpenShift-Based Applications - For applications exposed using OpenShift Routes.
• HTTP Scaling with Kubernetes Gateway API - For applications exposed via the Gateway API, Kedify utilizes its autowiring featurel.
• HTTP Scaling with Istio VirtualServices - This guide demonstrates how to scale applications within an Istio service mesh using VirtualService resources to manage HTTP traffic.
• HTTP Scaling with TLS for Ingress-Based Applications - For TLS workloads using Ingress, Kedify intercepts traffic and collects encrypted traffic metrics without interfering with your TLS setup.

Advanced Configuration

• Configure Waiting and Maintenance Pages for HTTP Scaler - Configuration options to set static pages when your application is going through cold-starts or needs temporarily re-route the traffic.
• Configure Envoy in the Kedify Proxy - How to configure Kedify Proxy Envoy for advanced use cases with custom Envoy config snippets.
• Kedify Proxy Performance Tuning - For applications that use Kedify Proxy to autoscale based on HTTP traffic.
• OpenTelemetry Tracing for Kedify Proxy & Interceptor - Enable OpenTelemetry tracing to get insights about HTTP calls handled by Kedify Proxy and HTTP Add-on interceptor components.

Custom Metrics and OTel Integration

• Scaling with Custom Metrics and OTel Collectors - Suitable for applications that expose custom metrics either as a scrapeable metrics endpoint or as a OTLP exporter. Example application is AI workload (namely vLLM with LLama 3.1)
• Custom Metrics with Prometheus Scaler & Migration to OTel Scaler - Setting up simple demo app that exposes application-specific metrics, Prometheus Scaler and then migrating everything to OTel Scaler.

Others

• PodResourceProfile shrink to zero - Shrink the last replica to less resources when ScaledObject is not active.