State-machine replication and consensus algorithms have been around for a long time and underpin several core-distributed primitives that power web-scale platforms, such as Google (Chubby, Spanner), Amazon Web Services and many others.
But do you have to work at this scale to benefit from such an architecture?
In recent years, at Adaptive we have built trading systems using the Raft consensus algorithm and embracing the concept of state-machine replication. This article covers the benefits (and some pitfalls) we’ve encountered, to help you decide whether such an architecture is right for you.
Is your company considering building a new system with this architecture, or are you a developer wanting to work on such projects? If so, we’d love to hear from you: