Provide alternative capacity for your critical system elements by duplicating those elements. Redirect users to the duplicate should the active element become unavailable.

This pattern must be applied equally to hardware and software. Software elements are, perhaps, more prone to failure than hardware because they are inherently more complex. They are also likely to be the elements that need to be taken out of service for maintenance (usually to have their functionality upgraded). However, this isn't to say that hardware is immune to failure or requires no maintenance. And there is no point having a single hardware element running an active-redundant pair of software elements. If the hardware fails, the system is down. Architecture introduces redundant switches and servers that will automatically swap in should the active one fail.