Crane: Cloud Resource Analytics and Economics


Architecture

Crane (FinOps Crane) is an open-source project which manages cloud resources on Kubernetes stack, it is inspired by FinOps concepts.
The goal of Crane is to provide a one-stop-shop project to help Kubernetes users to save cloud resource usage with a rich set of functionalities:

  • Resource Metrics Prediction based on monitoring data
  • Cost visibility including:
    • Cost allocation, cost and usage virtualization
    • Waste identification
    • Idle resource collection and reallocation
  • Usage & Cost Optimization including:
    • Enhanced scheduling which optimized for better resource utilization
    • Intelligent Scaling based on the prediction result
    • Cost Optimization based on better billing rate
  • QoS Ensurance based on Pod PriorityClass

crane-architecture

Features

Time Series Prediction


Knowing the future makes things easier for us.


Many businesses are naturally cyclical in time series, especially for those that directly or indirectly serve “people”. This periodicity is determined by the regularity of people‚Äôs daily activities. For example, people are accustomed to ordering take-out at noon and in the evenings; there are always traffic peaks in the morning and evening; even for services that don’t have such obvious patterns, such as searching, the amount of requests at night is much lower than that during business hours. For applications related to this kind of business, it is a natural idea to infer the next day’s metrics from the historical data of the past few days, or to infer the coming Monday’s access traffic from the data of last Monday. With predicted metrics or traffic patterns in the next 24 hours, we can better manage our application instances, stabilize our system, and meanwhile, reduce the cost.

Crane predictor fetches historical metric data for the monitoring system, such as Prometheus, and identifies the time series that are predictable, for example, system CPU load, memory footprint, application’s user traffic, etc. Then it outputs the prediction results, which can be consumed by other crane components, like Effective HorizontalPodAutoscaler and Analysis as a Service. It’s also straightforward to apply the prediction results in user applications.

Please see this document to learn more.

Effective HorizontalPodAutoscaler

EffectiveHorizontalPodAutoscaler helps you manage application scaling in an easy way. It is compatible with native HorizontalPodAutoscaler but extends more features.
EffectiveHorizontalPodAutoscaler supports prediction-driven autoscaling that supported by TimeSeriesPrediction. Users can forecast the incoming peak flow and scale up their application ahead, also users can know when the peak flow will end and scale down their application gracefully. Besides, EffectiveHorizontalPodAutoscaler also defines several scale strategies to support different scaling scenarios.

  • Reliability: Guarantee both scalability and availability
  • Responsiveness: Scale-up fast enough to successfully handle the increase in workload
  • Observability: Support Preview mod and automatic observe replicas

Please see this document to learn more.

Analysis as a Service

Analysis Service gives you recommendations about cost optimize. It scans your cluster resources such as Deployment, StatefulSet and provides a variety of strategies to analyze the resource then recommend how to optimize it. Analysis and Recommendation are CustomResourceDefinition that can integrate with your own systems.

Here we provide two Analysis Services:

  • ResourceRecommend: Recommend container requests & limit resources based on historic metrics.
  • Effective HPARecommend: Recommend which workloads are suitable for autoscaling and provide optimized configurations such as minReplicas, maxReplicas.

Application Ensurance

GitHub

View Github