Gossip, Membership, and Failure Detectors
Stub. Gossip (epidemic) protocols for disseminating information and maintaining membership, gossip-style and SWIM-style failure detectors, and a brief look at grid computing as an early large-scale distributed setting.
Two problems that recur in every large, dynamic distributed system: how do nodes disseminate information without a central broadcaster (gossip/ epidemic protocols — probabilistically reliable, scalable, and simple), and how does a node know which other nodes are still alive (failure detectors and membership protocols) — including the tension between detection speed and false-positive rate that gossip-style and SWIM-style detectors navigate differently. Grid computing is covered briefly as a contemporary large-scale use case. This machinery underlies both the consensus failure detectors (P, S) used later and the epidemic protocols peer-to-peer systems reuse for dissemination.
Outline (TODO — flesh out each)
- Gossip (epidemic) protocols
- Failure detectors — gossip-style failure detection
- SWIM and its predecessor
- Membership protocols
- Grid computing (brief)