Distributed systems fail differently from centralised ones — and their failure modes are often more manageable.
How Decentralisation Changes Failure
Centralised systems concentrate decision-making authority and operational capability in a small number of nodes. This concentration produces coordination efficiency under normal conditions — decisions can be made quickly by the central authority, resources can be allocated consistently, and the system can be optimised as a coherent whole. It also produces specific failure modes: the failure of a central node can disable the entire system, because the peripheral nodes lack the authority or capability to function independently when the central node is unavailable.
Decentralised systems distribute decision-making authority and operational capability across multiple nodes, each of which can function independently. This distribution produces coordination inefficiency under normal conditions — the absence of a central authority means that decisions must be negotiated, resources may be duplicated, and the system cannot be optimised as a whole. It also produces different failure modes: the failure of any individual node does not disable the system, because the other nodes continue to function and can compensate for the failed node's reduced contribution. The decentralised system fails partially and locally rather than completely and globally.
The Resilience Value of Decentralisation
The resilience value of decentralisation is highest when the relevant disruptions are geographic or node-specific — when they affect a subset of the system's components while leaving others unaffected. The pandemic that overwhelms a centralised healthcare system disrupts the entire system simultaneously; the same pandemic affecting a distributed network of healthcare providers disrupts some nodes while leaving others fully functional. The cyberattack that compromises a centralised data system compromises all data; the same attack on a distributed system compromises only the nodes it reaches.
The resilience value of decentralisation is lowest when the relevant disruptions are systemic — when they affect all nodes simultaneously regardless of geographic distribution. The financial panic that triggers a run on deposits affects all financial institutions simultaneously, regardless of whether the system is centralised or distributed. Against systemic disruptions, decentralisation provides no resilience advantage and may actually impair the coordinated response that the disruption requires.
Decentralisation is resilience against local failures. It is not resilience against systemic ones. Knowing which type of failure the institution faces is the precondition for knowing whether decentralisation is the appropriate resilience investment.
Discussion