Workflow

The primary idea behind Concentrated Dispersal is the realization that data loss is primarily caused by drive failures, rather than server failures. When components in a server fail, such as the power supply, CPU, memory, network card, or mother board, the server will suffer a temporary availability outage. However, the slices and the data held in the drives of the server are left intact. Therefore, while ensuring no more than one pillar is stored on a server serves to maximize availability, it does little to improve the reliability of data. What is critical for maximizing reliability is that no one drive should hold more than one pillar worth of data.

While it takes a very large number of nodes to use, for example, an IDA Width of 36, with one pillar per Slicestor, it does not take very many nodes to get 36 hard drives. Slicestors commonly have between 12 to 80 or more hard drives. Therefore, a very small number of nodes is sufficient to build a system that retains different pillars on different drives. By ensuring that drives are not responsible for more than one pillar, the maximum fault tolerance of (IDA Width - Threshold) number of disk failures can still be tolerated. This is the primary idea underlying Concentrated Dispersal.