State Database Replicas Examples

Desktop System: In a desktop system, you might have your root drive, plus a single controller with two external drives. Placing a few (approximately 3-5) state database replicas on each external drive, preferably on different slices, plus a few more (total of the number of the external drives, plus 1) on the root drive, would leave a relatively fault-tolerant configuration. Any single device failure (except the controller for the internal drives) would leave the system functional. A controller or drive failure on the root drive would leave the system non-functional, but it would be non-functional for other reasons as well. Short of creating mirrors for the rhot drive, this configuration is about the best that you can achieve with a small system configuration.

Small Server without Mirrored Root: A small server might have a root drive, plus external disk packs on separate controllers. A prudent configuration could place a state database replica on each of the disks within each of the disk packs, plus three on the root drive. This configuration would be resistant to failure of any single component except the root drive or controller. This differs from the previous configuration because of the two controllers for extra drives. With the replicas on a surviving controller, the number of state database replicas on the root drive needs only be enough to give a majority. The previous example requires that the root drive contain a majority.

Small Server with Mirrored Root: A small server with a mirrored root might have a root drive, plus external disk packs on separate controllers. Here too, a state database replica per drive, plus a couple of extras on the root drive would be appropriate. With a mirrored root (assuming the root drive is mirrored onto another drive and controller), there is no single point of failure, so the state database replicas need only be configured so that a single device or controller failure would disable less than 1/2 of the total number of controllers.