Difference between revisions of "Auto-scaling SIG/Theory of Auto-Scaling"
Joseph Davis (talk | contribs) (Initial framework) |
(No difference)
|
Revision as of 22:52, 15 May 2019
Contents
Theory of Auto-Scaling
General Description
<fill in>
Conceptual Diagram
If you prefer PlantUML
@startuml
cloud Cloud\n {
rectangle host as "Host" { } rectangle host2 as "Host" { agent VM agent VM2 as "VM" agent Container agent Container2 as "Container" }
}
agent MS as "Monitoring Service" agent DS as "Decision Services\n(Clustering,\nOptimization,\nRoot Cause)" agent Heat as "Orchestration \nEngine"
host -down-> MS VM -down-> MS Container -down-> MS : "Metric \nSamples"
MS -down-> DS : "Alarms" MS -down-> Heat : "Alarms"
DS -right-> Heat : "Scaling Commands"
Heat -up-> host : "Orchestration" Heat -up-> VM2 : "Orchestration" Heat -up-> Container2 : "Orchestration"
@enduml
Considerations and Guidelines
- Monitoring takes resources, plan accordingly
- Avoid scaling too quickly or too often
- Don't expect instantaneous scaling (see above)
- Be aware of where the logic for scaling is (alarm thresholds, decision services)