Kedify Architecture

Kedify extends the standard Kubernetes autoscaling stack instead of replacing it. KEDA still evaluates events and custom metrics, Kubernetes HPA still applies replica changes, and Kedify adds the missing pieces teams usually need in production: managed KEDA operations, HTTP autoscaling, richer observability, and centralized multi-cluster control.

How Kedify, KEDA, and HPA Fit Together

At a high level, the control loop works like this:

1. Your workload exposes demand through an event source, HTTP traffic pattern, or custom metric.
2. KEDA or a Kedify scaler translates that demand into autoscaling metrics.
3. Kubernetes HPA uses those metrics to scale replicas up or down.
4. Kedify provides the management, telemetry, policy, and advanced scaler capabilities around that loop.

This means Kedify is compatible with the core KEDA model while improving how teams operate autoscaling at scale.

Core Components

• Kedify Agent: The Kedify Agent is responsible for managing the installation and lifecycle of KEDA and the HTTP Scaler. It handles telemetry collection, providing users with valuable insights into their configurations. The agent establishes a secure, outbound gRPC connection to the Kedify Service, ensuring secure communication.

• Kedify Dashboard: The Dashboard facilitates the management of installations and offers insights based on the telemetry data collected by the Kedify Agent. It serves as the central interface for monitoring and configuring Kedify-managed resources.

• Kubernetes Integration: The Kedify Agent is deployed as a standard Kubernetes deployment. It manages Kedify custom resources, providing autowiring and comprehensive management of the HTTP Scaler. For more information on custom resources, refer to Kedify Custom Resource Definitions.

• Telemetry Collection: The Kedify Agent includes a telemetry collector that scrapes metrics from both KEDA and the HTTP Scaler. These metrics are crucial for providing real-time insights and for monitoring the performance and scaling behaviors of your applications.

Where Kedify Extends Upstream KEDA

Kedify is most useful when you need more than raw event-driven autoscaling:

• HTTP Scaler for HTTP-aware scaling, traffic autowiring, and scale-to-zero behavior
• OTel Scaler for autoscaling from OpenTelemetry metrics
• Predictive Scaler for proactive scaling before demand spikes arrive
• Multi-Cluster Scaling and Scaling Groups for larger fleet operations

By combining these components, Kedify gives teams a more complete Kubernetes autoscaling architecture without forcing them away from KEDA or HPA.

For more detailed information about each component, please refer to their respective documentation:

• Kedify Agent
• Kedify Dashboard
• Kedify Custom Resource Definitions