Important terms

These terms are used throughout this document.

Node

A generic term for a component that encompasses all types of that component: appliance, virtual appliance, application, or Docker container.

Device Set

A set of Slicestor® Nodes of a multiple of the Least Common Multiple of an IDA Width that either forms a new Storage Pool or expands an original Storage Pool. The amount of Slicestor® Nodes in a Device Set must be an integer multiple of the Least Common Multiple of all the IDA widths for Vaults the Device Set serves.

Storage Pool

A grouping of one or more Device Sets which together provide the physical storage resources for one or more vaults.

Source Slicestor® Node

A Slicestor® Node that sends Slices during reallocation to its destination peer in the new Device Set. During reallocation, this node serves as a proxy for all operations to its destination peer so that reallocated Slices remain visible and accessible to Accesser® Nodes and rebuild processes.

Destination Slicestor® Node

A Slicestor® Node that receives Slices that are sent during reallocation from its corresponding source peer in the current Device Set. During reallocation, this node handles operations that are proxied by source nodes.

Weighted Rendezvous Hash (WRH)

The algorithm that is used to select which Device Set manages a given name. This algorithm minimizes reallocation upon any change by deterministically mapping an object to its Device Set. The algorithm can handle changes to the underlying system efficiently.

Slicestor® Node Proxying

As Slices are reallocated, the source Slicestor® Node tracks which Slices are successfully moved to the destination Slicestor® Node. Since the source Slicestor® Node no longer holds the Slice after it is reallocated, and because Accesser® Nodes do not communicate to destination Slicestor® Nodes during reallocation, the source Slicestor® Node must proxy the read, write, and list requests it receives to be serviced by the destination Slicestor® Node. When the destination Slicestor® Node responds to its request, the source Slicestor® Node returns the destination Slicestor® Node response to the client.

Throughput

The total inbound and outbound rate that the Slicestor® Node can process, usually expressed in Megabytes per second (MBps).

OPS

The number of possible operations (reads / writes) per second.

Latency

The time to process an HTTP request.

Concentrated dispersal

A technique whereby each Slicestor device may be responsible for multiple slices of a given object stored in a system. For instance, a system of three Slicestor devices can have an IDA width of 36, using twelve slices for each Slicestor. To use a Concentrated Dispersal vault, a device set must be created using between three and six Slicestor devices.