Reduce the autoscaling and metric collection window (linux and windows)
Currently, the metrics are collected every 15 minutes when scaling based on CPU. We can reduce the scaling frequency down to 15 minutes causing scaling operations to occur up to 30 minutes after the threshold is met. I'd like to see this time decreased significantly on Linux and windows VM's. In live hosting environments, we should be able to scale in minutes if not seconds based on cpu threshold.
Thank you for your feedback and patience. We are working on significantly improving the metrics, alerts and autoscale pipelines/jobs and this will reduce the end-to-end latency.
- Ashwin Kamath, PM, Azure Insights