| Building online services that are both highly available and correct is challenging. Byzantine fault tolerance (BFT), a technique based on state machine replication, is the only known general technique that can mask arbitrary failures, including crashes, malicious attacks, and software errors. Thus, the behavior of a service employing BFT is indistinguishable from a service running on a correct server.;This dissertation presents three practical BFT protocols, hBFT, BChain, and ByzID. Each protocol takes a different approach enhance the practicality of existing practical BFT protocols under certain network conditions and threat models. hBFT moves some jobs to the clients with minimum cost. The protocol is much simplified while faulty clients are tolerated. BChain uses chain replication while faulty replicas are diagnosed and eventually reconfigured. ByzID uses intrusion detection methods to build a Byzantine failure detector. Faulty replicas are detected immediately and performance attack can be perfectly handled. In the end, we present P2S, a general framework of adapting existing fault tolerance techniques to pub/sub, with the aim of reducing the burden of proving the correctness of implementation. The experimentation results validate all the work, showing different degree of performance improvement over traditional protocols. |
