The solution is to employ a software system to perform scaling automatically based on parameters such as number of requests, CPU utilization, memory utilization, and buffer size (queue size), and so on to improve the efficiency of the system and to optimize operational expense.

Auto-scaling is an elastic process which will provision resources based on the work load and a preconfigured collection of rules specifying the thresholds, scaling range, scaling factors, and so on. The system should be capable of scaling the resources vertically, by redeploying the service on a more capable hardware or, scaling horizontally, by provisioning additional instances of the services.

The following diagram illustrates an auto-scaling system in action:

Azure Service Fabric clusters are built on top of virtual machine scale sets and can be scaled either manually or with auto-scale rules. You can read more about building auto-scaling rules at: https://docs.microsoft.com/en-us/azure/service-fabric/service-fabric-cluster-programmatic-scaling.