Chapter 10, Modern Design Patterns for Scalability

1. No; virtual and physical networks are created with a finite amount of resources and cannot be autoscaled. Monitoring conditions related to network traffic can be a means of triggering an autoscaling event to better handle an influx of traffic.
2. CPU and memory tend to be the easiest resources to adjust and will generally be seen as the targets for monitoring usage or over-usage.
3. Out of the box, HPAs support evaluating thresholds on CPU and memory usage relative to the pod(s) assigned to the application. While other types of metrics can be leveraged, from monitoring other cluster components or even outside hosted systems, the recommendation is to stick with usage patterns with HPAs that are supported by default.
4. In the short term, yes. Longer term, even increasing the number of resources for the cluster nodes can become less maintainable and take away from your ability to separate workloads efficiently. Having a separation...