AI-based predictive autoscaler for Kubernetes

PredictKube AI-based KEDA scaler
Compatible with

Make your Kubernetes autoscaling proactivе

From reactive scaling to proactive
With PredictKube, you’ll be able to finish autoscaling before the load rise thanks to predictions made by our AI model.
Kick-off start
Our AI model can start working with the data for 2 weeks to provide you with reliable prediction and autoscaling.
Less manual scale, more automation
The predictive Keda scaler named PredictKube helps you to minimize time-wasting on manual setup of autoscaling and gives you an automated performance.
PredictKube KEDA scaler for Kubernetes

Ready to kick off now?

Install PredictKube KEDA Scaler in a few steps:
Add Helm repo
01
Copy
helm repo add kedacore https://kedacore.github.io/charts
Update Helm repo
02
Copy
helm repo update
Install keda Helm chart
03
Copy
kubectl create namespace keda
helm install keda kedacore/keda --namespace keda
Create PredictKube Credentials secret
04
Copy
API_KEY="<change-me>"
kubectl create secret generic predictkube-secrets --from-literal=apiKey=${API_KEY}
To make our AI model access your data and make a prediction based on it, please use the API key we'll send to your e-mail. Review our Privacy Policy to see how your data circulates inside PredictKube.
Configure Predict Autoscaling
05
Copy
tee scaleobject.yaml << EOF
apiVersion: keda.sh/v1alpha1
kind: TriggerAuthentication
metadata:
  name: keda-trigger-auth-predictkube-secret
spec:
  secretTargetRef:
  - parameter: apiKey
    name: predictkube-secrets
    key: apiKey
---
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: example-app-scaler
spec:
  scaleTargetRef:
    name: example-app
  pollingInterval: 60
  cooldownPeriod: 300
  minReplicaCount: 3
  maxReplicaCount: 50
  triggers:
  - type: predictkube
    metadata:
      predictHorizon: "2h"
      historyTimeWindow: "7d"  # We recommend using a minimum of a 7-14 day time window as historical data
      prometheusAddress: http://kube-prometheus-stack-prometheus.monitoring:9090
      query: sum(irate(http_requests_total{pod=~"example-app-.*"}[2m]))
      queryStep: "2m" # Note: query step duration for range prometheus queries
      threshold: '2000' # Value to start scaling for
    authenticationRef:
      name: keda-trigger-auth-predictkube-secret
EOF

Under the hood: Tools inside

We made our KEDA scaler out of the top-notch technologies
available for Kubernetes and AI

Input the data for 1+ week and get proactive autoscaling
up to 6 hours horizon based on AI prediction

number of nodes
monitoring time
PredictKube KEDA scaler prediction graph
Predicted nodes
RPS Predicted
RPS Actual

Our colleagues, clients, partners

We have the pleasure of being good friends and problem-solvers for these projects:

FAQ

How can I install PredictKube for my project?
It’s quite easy: use our quick start tutorial or check the official documentation of Keda scalers first. If you have any questions regarding the installation of the scaler, you can contact us directly or ask the community.
What does proactive scaling mean?
Proactive scaling means “in advance,” and it’s important for the projects that may need time to deploy more resources for their needs. The right time for scaling is selected by our trained AI model that analyzes your historical data and can utilize the data of custom and public business metrics that can affect the traffic load.
What part of the scaler is responsible for prediction functions — the KEDA-part or another?
All prediction functions are going on the SaaS server, through the API of PredictKube. If you want to know more about how we’ve done the “juicy” AI part of PredictKube, let us know.
Will PredictKube be free constantly?
We’ll support the free access to API in general with all basic features available to provide autoscaling possibilities to anyone who’s interested. There are plenty of options we’re currently researching, so we hope to enhance the PredictKube even further.
How can I contact your team?
The best way to address your questions and concerns to our team is to use a feedback form on our website. Also, you can write to the Dysnix team; we’re open to new acquaintances. We’re looking forward to hearing from you, be sure you use your corporate email, so we’ll prioritize your request.