Now it is hardly to find an architecture that does not claim to be fault tolerant or even provides this feature. Let us take a look for example at fault-tolerance ideas to world economy…well, yeah, it’s too late for talks. Let us talk about fault tolerance in ActiveRoles instead.
In the same way Active Directory is not fault tolerant with a single domain controller ActiveRoles Server is not fault tolerant when a single server running the ActiveRoles Server administration service is deployed. It is critical that at least two servers running the Administration Service be deployed to have a fault tolerant deployment.
An additional benefit is that both the MMC console and the web interfaces will fail-over to a new Administration Service if the first service becomes unresponsive. The user experience is slightly different depending on what interface is used when the administration service crashes. Within the MMC console the user will notice the administration service has stopped and will only have to use the Connect command to automatically connect to the next available service. Users of the web interface will have a more seamless transition as the web site fails over automatically to the next Administration Service. One important item to note is that automatic failover only works if the option ‘Any available Administration Service’ was selected during the Web Interface setup.
WRT database, ActiveRoles supports SQL cluster technology for database high availability. Database mirroring is to come in the nearest future. Note, that if you have SQL Server and ARS installed at different servers (and if stars are in right positions on the sky) existing MMC and Web Interface sessions may still may available (with certain functionality disabled) and AD changes will be applied to AD successfully.
Keep in mind that even in the worse case scenario where all ActiveRoles Servers fail, Active Directory will continue to function normally. The only result of a complete failure is that day-to-day administration or help desk functions may be interrupted until a server is brought back online.
Also, check out this wiki link that can be extremely useful in fault tolerance deployment