Cloud Metrics

Aus SDQ-Wiki
Wechseln zu: Navigation, Suche
Palladio Addon (Liste)
Name Cloud Metrics
Contacts Steffen Becker (https://www.iste.uni-stuttgart.de/rss/)
State Incubation
Is Stand-alone Analysis? No
Extends Analyses SimuLizar
Extends Metamodels
Update Site https://sdqweb.ipd.kit.edu/eclipse/simulizar/nightly/

Short Summary

Cloud metrics are part of SimuLizar and allow the user to analyse scalability, elasticity, and efficiency.

Documentation

Motivation

Cloud Metrics cover analyses of the capabilities of one of the most essential parts of a cloud computing system: the elasticity management. Most of the metrics try to capture to some degree elasticity properties, i.e., to which degree does the elasticity management match the provisioned resources to the demanded ones without violating the end-to-end SLOs of the overall system. A counter-player of these is efficiency: the more elastic the system is, the less efficient it usually becomes.

Metrics

The following discusses the available metrics distinguished along the categories of scalability, elasticity and efficiency metrics.

Scalability

SimuLizar can perform transient analyses of adaptation behavior to predict the following metrics:

  1. Scalability Range: The load range in which the system under a fixed work is operating without violating any of its SLOs

Elasticity

SimuLizar can perform transient analyses of adaptation behavior to predict the following metrics:

  1. MeanTimeToQualityRepair: The time it takes for the elasticity management to deal with a temporary phase of violating SLOs
  2. SLOViolationsOverTime: The number of overall, end-to-end system runs which experience an SLO violation. In SimuLizar you have to model the SLOs explicity for this metric to be computable.

Efficiency

SimuLizar can perform transient analyses of adaptation behavior to predict the following metrics:

  1. NumberOfResourceContainersOverTime: The number of resource containers provisioned over time. This is useful to evaluate horizontal scale-out scenarios.
  2. OverallCost: Based on the previous metric and an additional cost model, SimuLizar predicts the overall cost of the analyzed scenario.

Further Reading

Detailed definitions, the analysis process, and descriptions of the analyzable metrics can be found in the CloudScale book published by Springer.