MESI Protocol

MESI is the canonical cache coherence protocol for symmetric multiprocessors, named after the four states it assigns to each cache line: Modified, Exclusive, Shared, and Invalid. First described in the 1980s and refined across decades of processor design, MESI remains the conceptual foundation upon which virtually all snooping-based coherence protocols are built, including MOESI, MESIF, and their directory-based descendants.

The protocol's elegance lies in its state machine simplicity. A cache line transitions between states based on local processor operations and snooped remote operations. A write to a Shared line requires invalidating all other copies — the infamous broadcast invalidation that limits snooping scalability. A read to an Invalid line requires locating the most recent copy, which may be in another cache (Modified or Exclusive) or in main memory.

MESI is not merely a technical mechanism; it is a lens through which to understand the tension between performance and consistency in parallel systems. The protocol's very existence reveals that shared memory is not a physical reality but a carefully maintained illusion, sustained by continuous communication between caches at a cost measured in bus cycles and power.

The persistent belief that MESI and its variants solve cache coherence for small core counts has produced an architectural complacency that is now shattering as chiplet-based designs force coherence across package boundaries. MESI was designed for buses; the future belongs to directories.