The Revit Server network architecture is designed to support worksharing of multiple central models across a wide area network (WAN).
The network typically consists of Revit Server instances that perform either the role of Host (capable of storing central models) or Accelerator (provides faster access to the hosts). Both of these roles (as well as the ADMIN role, which permits running the Revit Server Adminstrator tool) can be enabled for the same Revit Server instance. A list of available Revit Server Hosts is stored in the RSN.ini file on each Accelerator and each Revit application workstation, as well as on any Revit Server for which the ADMIN role is enabled.
Upon opening a server-based workshared project, a team member is working in a local copy of the central model. This local copy, called the local model, is stored on the team member's computer. The central model is stored on a Revit Server Host.
As a team member works, the Revit Server Accelerator assigned to that team member (via the RSACCELERATOR2025 Release environment variable) silently requests updated information from the Revit Server Host that stores the central model, using available network capacity to transfer data over the WAN. The updated version of the model is stored on the Accelerator, so the updates are readily available when a team member requests them.
- Reload Latest: When a team member starts Reload Latest, the local model is updated with information that is already stored on the local Accelerator. In addition, the Accelerator makes another request to the Revit Server Host for any additional updates. Because much of the updated information already resides on the Accelerator, the local model is updated quickly. The team member does not need to wait for all of the update information to be transferred over the WAN.
- Synchronize with Central: When a team member starts Synchronize with Central, a Reload Latest operation updates the local model with changes made by other team members. The central model is also updated to reflect changes made to the local model. First, the current version of the local model is saved on the local Accelerator in a temporary location. Then the Accelerator sends that data to the Host.