Resumen
The STAR protocol is introduced, which solves three problems with MPI, a well known secur- ity problem, and three exascale communication problems. Optical implementations are developed compatible with 100 Gbit/sec Ethernet. Automatic fault resilience mechanisms are discussed, which improve HPC quality of service, and meet the exascale reliability and resilience challenges. The bandwidth problem for exascale computers interfacing with data centers is solved. The STAR protocol is combined with the SiMulPro core architecture (discussed in another article of this issue). The combination enables data centers, handheld computers, networked sensors, and super- computers, to be invulnerable to memory fault injection of viruses and rootkits.