Автор: Avesha, Inc.
Smart Scaler guarantees SLAs and reduces cloud costs by autonomous autoscaling of K8s with AI and RL
Smart Scaler uses Reinforcement Learning (RL) to guarantee application SLA (for example: response time, error rate) and reduce cloud costs by precisely scaling the K8s application resource needs just in time. It will:
- Extract application and infrastructure performance data from the cluster’s monitoring data sources (e.g., Prometheus, Datadog, NewRelic, etc).
- Use historical application performance data to predict pod capacity requirements and the number of pods needed for a given load.
- Build AI/ML/RL models of application behavior and traffic loading patterns to achieve the application SLA by the collaboration of the microservice scaling models, taking into account the service graph.
- Apply these models toward monitoring the application and increasing or decreasing resources in anticipation of the application's need.
- Free up top kubernetes, devops, engineering talent from worrying about SLA and eliminate performance tuning headaches and overprovisioning.
- Smart Scaler deals with event-based traffic spikes by autonomous autoscaling.