Determine slice allocation

Edit online

The Slicestor® Nodes use the Weighted Rendezvous Hash algorithm to determine where to move the Slices. It is determined by the utilization of each Device Set.

Example 1. How the WRH algorithm works

The WRH algorithm takes a name and the list of Device Sets and the capacities as inputs and returns the Device Set that should store that object.

Consider a Storage Pool with three Device Sets: A (10 PB), B (25 PB) and C (20 PB). The total of all the weights are 10 PB + 25 PB + 20 PB = 55 PB.

The probability that any data source name is mapped to the Device Sets is:

  1. (10 PB / 55 PB) = 18.18%
  2. (25 PB / 55 PB) = 45.45%
  3. (20 PB / 55 PB) = 36.36%
Note: When new data is written, all the Device Sets fill up at a rate that equalizes their storage usage as a percentage of the total storage capacity. In the stable state, all Device Sets (and all Slicestor® Nodes within them) are used equally.

Any time the capacity of a Device Set changes or a Device Set is added, some Device Sets have more data than others. This algorithm is reapplied to all names to store the Slices in their proper locations and restore equilibrium.