Scaling Out

When load outgrows a system, there are two ways to add capacity: give a server a bigger machine, or run more servers side by side. You will see each one applied, a database moved to a larger machine and an application tier spread across ten small ones, and then weigh them against each other, including against the size of the team that has to run the result. The topic then covers what scaling out demands first, that no user's session lives in one server's memory, and what autoscaling does and cannot do when a spike arrives faster than a server starts. Splitting a database across machines, sharding, belongs to the subject on distributed data and is named here only as the step beyond.