19.2 How to Scale and Stabilize Workloads in Enterprise-Level Organisations

• Vertical Scaling Limitations and Horizontal Scaling to the rescue

  • Before attempting horizontal scaling, it is usually always good to exhaust the portion of vertical scaling where costs climb linearly with computation.

• High Availability is achieved by

  • Using Load balancers in Horizontal scaling

    • Imagine a restaurant with one chef that is trying to serve meals to a hundred customers at once.

      • The chef can only cook one meal at a time, so the customers are getting frustrated with long wait times and some of them are leaving.This is because the chef is overloaded and cannot handle the intense workload.

      • Now, imagine the restaurant hires three more chefs, all with the same skills and equipment as the first chef. The restaurant invests in a system where each chef has a buzzer that goes off when there are new orders waiting for them. The buzzer ensures that each chef knows exactly when they need to start cooking a new meal, and they do not get overwhelmed by too many orders at once.

      • In this analogy, the chefs represent servers or clusters, and the buzzer represents a load balancer.

  • Maintaining a backup of the load balancer and the database/NAS.

• Adding Autoscaling (i.e. you save money when traffic is low)
  • Many autoscaling frameworks these days assume that applications are mostly stateless.