Hands-On Software Architecture with Golang

Storage system-centric consistency model

This section discusses consistency from the perspective of the storage system. Generally, the storage system has multiple replicas to allow for resilience (redundancy) and scalability. As part of accepting a write, the storage systems needs to decide how many of these replicas have to be updated before the write can be acknowledged to the client.

Let the terms be as follows:

• N: The number of nodes in the storage system
• W: The number of replicas that are updated before the write is acknowledged to the client
• R: The number of replicas that are contacted for a read (for a quorum, see the following)

If W + R > N, then it will never happen that processes see inconsistent data, thereby guaranteeing strong ...