Im­ple­men­ta­ti­on and Em­pi­ri­cal Eva­lua­ti­on of a Light­weight Me­di­an-Ru­le Pro­to­col for State Ma­chi­ne Re­p­li­ca­ti­on

Replication exists to eliminate single points of failure. Yet classical state
machine replication routes every command through a leader, at once a
throughput bottleneck and the system's most exposed attack target. "A
Lightweight Approach for State Machine Replication" (Cachin, Dou, Scheideler,
Schneider, 2025) instead adapts a simple median rule from stabilizing
consensus into a leaderless, gossip-based protocol with compact server state.
The proven guarantees are strong: agreement in O(log n) rounds w.h.p.,
liveness as long as DoS (blocking) attacks remain below a threshold, bounded
storage, and rapid recovery once an attack subsides. They are also purely
theoretical. Since the protocol has never been implemented, the constants
behind the asymptotic bounds are unknown, as is the behavior between the
liveness threshold and the death-spiral threshold. This thesis will
contribute the first implementation and an empirical evaluation, based on
deterministic, exactly reproducible round-based simulation of networks of up
to thousands of servers. The evaluation proceeds in two stages: it first
validates the implementation against the proven guarantees, and only then
measures what the theory leaves open. The talk presents the protocol, the
evaluation design, and the work plan.