Eleven Fallacies of Distributed Computing

Originally identified by Peter Deutsch (1994) and colleagues at Sun Microsystems; expanded over time as distributed systems practice evolved.

#	Fallacy	Why It’s Wrong	Mitigation
1	The network is reliable	Packets get lost, connections drop, switches crash	Retries with exponential backoff, timeouts on every call, circuit breakers
2	Latency is zero	A network call is a million times slower than a local call	Minimize round trips, batch operations, use async where possible
3	Bandwidth is infinite	Bandwidth is limited, shared, and costs money in the cloud	Watch payload sizes, compress data, paginate large results
4	The network is secure	Traffic passes through infrastructure you don’t control	TLS in transit, authenticate every request, authorize every operation
5	Topology doesn’t change	Services move, scale, fail, and get replaced continuously	Use service discovery, design for dynamic topology, health checks
6	There is one administrator	Different teams own different parts of the infrastructure	Document dependencies, communicate changes early, version APIs
7	Transport cost is zero	Bandwidth, infrastructure, and serialization all cost money and CPU	Factor transport cost into architecture decisions, measure actual cost
8	The network is homogeneous	Systems run on mixed hardware, OS, and software versions	Use standard protocols, version APIs, test across environments
9	Versioning is easy	Old and new service versions run simultaneously during rollout	Plan backward compatibility, support multiple versions during transition
10	Compensating transactions always work	Some operations can’t be cleanly undone	Design for idempotency, understand saga limitations before relying on them
11	Observability is optional	Debugging a distributed system blind is nearly impossible	Invest in tracing, structured logging, and metrics from day one

Related guides