stepscale AI
AI-powered autoscaling that learns from your workload and optimizes itself.
Overview
stepscale AI is an intelligent autoscaling platform that uses historical metrics to automatically tune your scaling configurations. Instead of manually setting thresholds, min/max tasks, and scaling ratios, stepscale AI analyzes your actual workload patterns and continuously optimizes these values for you.
The result: lower cloud costs, faster response to traffic changes, and zero manual tuning.
How It Works
stepscale AI operates as a tuning layer that wraps around your existing reactive autoscaler:
- Your reactive scaler (Fast Autoscaler, Kubernetes HPA, or native AWS autoscaling) handles real-time scaling decisions every 1-2 minutes
- stepscale AI collects metrics - queue depth, task count, processing rates, timestamps - building a historical picture of your workload
- Periodically, the AI analyzes patterns - peak hours, idle periods, processing time variations, traffic correlations
- Optimized configuration values are generated - updated thresholds, min/max tasks, and scaling ratios
- Your reactive scaler picks up the new config and operates with optimized parameters
The AI runs infrequently (a few times per day), keeping costs minimal while delivering continuous optimization.
Key Features
Auto-Tuning
Automatically adjusts scaling thresholds, min/max task counts, and tasks-per-message ratios based on observed workload patterns. No more guessing at configuration values.
Anomaly Detection
Identifies unusual traffic patterns - a spike at 3am looks different from your daily 9am rush. Different scaling strategies for anomalies vs normal traffic, with alerting integration.
Cost Optimization Insights
Actionable reports showing where you're over-provisioned, how much you could save, and before/after comparisons. The kind of data that justifies the tool to your engineering manager.
Multi-Platform Support
Works with both AWS ECS and Kubernetes environments. Start with one, expand to the other without changing your workflow.
Relationship to Fast Autoscaler
Fast Autoscaler is our free, open-source reactive scaling engine for ECS. It handles the real-time scaling decisions based on queue metrics.
stepscale AI enhances Fast Autoscaler (and other scalers) by adding the intelligent tuning layer on top. You can use Fast Autoscaler standalone, or pair it with stepscale AI for optimal performance.
| Fast Autoscaler | stepscale AI | |
|---|---|---|
| Reactive scaling | Yes | Yes (via Fast Autoscaler or K8s HPA) |
| Manual configuration | Yes | Auto-tuned |
| Anomaly detection | - | Yes |
| Cost insights | - | Yes |
| Kubernetes support | - | Yes |
| Price | Free & open source | Pro |
Get Early Access
stepscale AI is currently in development. Contact us to request early access and help shape the product.