The following post comes from Mike in our QA group for Revit Server.
Revit Server 2013 supports the deployment of multiple host servers throughout the Revit Server Network. Some users believe that these host servers function as mirrors to each other, such that a model stored on one host would be transparently mirrored to another, akin to a RAID 1 disk array. However this is not the case. A more accurate mental model is that of multiple file shares on a network. Just as a network can be configured with several shared folders for the storage and organization of shared content, the Revit Server Network can be configured with several hosts for the storage and organization of Revit models. Each host is independent of every other, and there is no mirroring or replication between host servers.
Using multiple host servers affords several benefits:
- Improved fault tolerance. For instance, if one host server is disrupted by a power outage, the content at the other host servers is still available.
- Opportunities for performance optimization. Models can be stored and moved throughout the network to keep them as close to the "center of activity" for a project as possible. Furthermore, server role configurations can be dynamically adjusted to accommodate project needs as they evolve.
- Improved utilization of network resources. Using multiple host servers can help distribute and balance the network traffic required for effective collaboration.
When multiple hosts are deployed, each accelerator has the ability to cache content from any host server in use. If a particular instance of Revit Server has host and accelerator roles enabled, the accelerator will not cache the content stored on that server as a host (there is no need to; it is already there) but it will cache content stored on any other host.
The Revit Server wiki help has a supplemental video on this topic, located here: