vamana

Vamana is an auto scalar that helps you scale your software or hardware based on application demand. Making Elasticity truely elastic.

How does it work?

Vamana has an Autoscaling algorithm which identifies your application load based on a Demand-Supply model and determine if it has to Upscale / Downscale. In order to integrate your application with Vamana it needs to implement what are called as a "Collector" and a "Scalar".

Collector

case class ResourceStat(demand: Demand, supply: Supply, timestamp: Long)

trait Collector {
  def getStats: ResourceStat
}

Collector is used for collecting stats about the application in a Demand-Supply fashion. Demand represents how much unit of work is being done / still pending while Supply represents how much capacity is present in the application.

Scalar

Scalar is used to calculate given a normalised demand and supply. How much nodes are required to satisfy the demand / how much nodes to be downscaled.

case class ScaleUnit(numberOfNodes: Int)

abstract class Scalar(cluster: RunningCluster) {
  def scaleUnit(normalizedStat: ResourceStat): ScaleUnit

  def downscaleCandidates(number: Int): List[String]
}

Scalar is also used to determine which are all the nodes which has to be downscaled. It could be based on oldestRunning / latestRunning / machineWithLowestDiskSpace / machineWithLowestMemory, etc.

API

Vamana runs as a simple HTTP service that exposes the following endpoints

You can find a Postman Collection under the root of the repo as postmanCollection.json.

Built With

• jclouds
• python
• scala