Horizontal Scaling

1. The Incident

The store has operated its application tier on one increasingly large server for several years. It eventually reaches the limits from the last lesson's discussion: the next size costs too much, and one host failure recently stopped checkout for 40 minutes. The overdue solution is to run two servers behind a simple traffic distributor that sends half of the requests to each.

The deployment happens on Tuesday morning before several problems appear that afternoon:

• **Users are randomly logged out