Modifying an existing CoreMedia installation, as described in the previous sections, results in a period during which the system is unavailable. The duration of that period is expected to be small, typically a few minutes, if the migration is properly planned and tested. After the migration the server and CAE caches are refilled, which can be a considerable time, too, depending on the actual load profile. If the overall availability expectations are not fulfilled using this scheme, it is suggested to use multiple Replication Live Servers of which only one Replication Live Server is allowed to replicate the repository structure change at a time, whereas the other Replication Live Servers must have stopped replication. Only the CAE attached to the Live Server that is currently replicating have to be switched off and replaced by new CAE that can cope with the new repository structure. This way, most of the CAEs will be up at any given time. This is complex and not normally necessary, though. It also requires much more project-specific planning, so that a detailed description is omitted here. In many cases the downtime during steps 9 to 12 of Section 3.11.4.3, “Migrating the Content Servers” can be bridged by setting up a web server delivering a set of static pregenerated pages, if that is needed.
 |  |  |
 |