New capabilities

Storage Pool Expansion offers a new method to increase storage capacity of original Vaults.

A new set of Slicestor® Nodes (a Device Set) is added to an original Storage Pool. After the new Device Set is added, a portion of the data stored on the original Device Set is reallocated to the new Device Set. This increases the storage and throughput capacity of that Storage Pool by an amount corresponding to the new Device Set.

Prior to version 3.6.0, storage capacity could be increased in three ways:

1. Adding drives to partially populated Slicestor® Nodes
2. Merging an original Storage Pool to a newly added Storage Pool
3. Creating additional Storage Pools

Merging Storage Pools created generations of Storage Pools. This process worked with Simple Object deployments but not with Cloud Storage Object (CSO) deployments.

Simple Object methodology has the system create the object name. This name can be encoded to indicate deterministically on which generation an object resides. (A deterministic system can predict the output with 100 percent certainty.)

Named Object methodology uses an object name that the end user provided. It is not possible to encode generation information to this object. Without some other mechanism, Named Objects can store and access objects deterministically only on the first generation.

As metadata was always stored to the first generation of a Vault using Named Object, when more and more objects are stored, the first generation may fill eventually and be unable to accept new metadata, index, synced or embedded objects (making the system unable to accept new writes even if other generations have available capacity). Once the first generation is full, the Vault would effectively become read only.

Storage Pool Expansion introduces four changes:

1. Moving away from multi-generational storage to multiple Device Sets
2. Using a new algorithm to make deterministic actions in storing objects on one of many possible Device Sets
3. Older device sets can be replaced with a new storage pool set.
4. More capacity can be added to a partially populated device set

Replacement is intended to remove old device sets that contain outdated hardware. Existing device sets can be resized to add more capacity.

The weighting used in this new algorithm is based on the relative capacity of each Device Set. This has the added benefit of ensuring that all Device Sets fill at the same rate.

Vaults that use Named-Object interfaces such as CSO- or OpenStack-compatible APIs can now be scaled up by adding one or more Device Sets to increase capacity of an original Storage Pool.