Real-World Network Behavior and Datacenter Failures
Stub. Empirical structure of real networks (topology, measurement) as distinct from idealized models, and case studies of real datacenter disasters — what actually causes large-scale outages in practice, as opposed to the fault models assumed by the theory.
A reality check against the rest of the course: the theoretical models (crash/omission/Byzantine faults, clean synchronous/asynchronous boundaries) are approximations, and this note collects the empirical evidence for how real systems actually behave and fail. Covers measured properties of real network topologies (which rarely look like the clean graphs used in the algorithmic notes) and worked case studies of actual datacenter outages, drawing out the recurring root causes (configuration error, correlated/cascading failure, insufficient testing of the failure path itself) that no single fault-tolerance technique fully addresses.
Outline (TODO — flesh out each)
- Structure of real-world networks — measured topology vs. idealized models
- Case studies: datacenter disasters
- Recurring root causes across incidents