Within any enterprise-wide system, if one node fails, you have a problem! Within a hierachical model, this could effect the mail delivery for the entire company so adding fault tolerance to your company-wide network becomes important.

Most of the mail that users will send will tend to be to users on their own LAN. For this reason, it's a good idea to limit the functions of the post offices to either working as a master post office or to supporting local users. This way, if a post office crashes, it won't effect two functions within the WAN. For this reason, try and keep users to the slave post offices at the lowest levels of the hierachy model, and the master post offices need only then concern themselves with delivering mail.

Another strategy is to build in redundancy within the network. This can take two forms: either redundant links or a redundant master post office. The latter is often the most effective, but is generally more expensive.

The spare master post office is keept up to date with a full user list and address directory. It also has a set of links to the slaves and the main master post office, but serves no purpose unless their is a fault. In this case, you can switch it on and it can immediately start routing mail between the slaves.