Check out the new USENIX Web site. Next: Acknowledgments Up: Contents Previous: Related Work

9 Conclusions

This paper has described a new state-machine replication algorithm that is able to tolerate Byzantine faults and can be used in practice: it is the first to work correctly in an asynchronous system like the Internet and it improves the performance of previous algorithms by more than an order of magnitude.

The paper also described BFS, a Byzantine-fault-tolerant implementation of NFS. BFS demonstrates that it is possible to use our algorithm to implement real services with performance close to that of an unreplicated service - the performance of BFS is only 3% worse than that of the standard NFS implementation in Digital Unix. This good performance is due to a number of important optimizations, including replacing public-key signatures by vectors of message authentication codes, reducing the size and number of messages, and the incremental checkpoint-management techniques.

One reason why Byzantine-fault-tolerant algorithms will be important in the future is that they can allow systems to continue to work correctly even when there are software errors. Not all errors are survivable; our approach cannot mask a software error that occurs at all replicas. However, it can mask errors that occur independently at different replicas, including nondeterministic software errors, which are the most problematic and persistent errors since they are the hardest to detect. In fact, we encountered such a software bug while running our system, and our algorithm was able to continue running correctly in spite of it.

There is still much work to do on improving our system. One problem of special interest is reducing the amount of resources required to implement our algorithm. The number of replicas can be reduced by using f replicas as witnesses that are involved in the protocol only when some full replica fails. We also believe that it is possible to reduce the number of copies of the state to f+1 but the details remain to be worked out.


Miguel Castro and Barbara Liskov,  "Practical Byzantine Fault Tolerance", in Proceedings of the Third Symposium on Operating Systems Design and Implementation, New Orleans, USA, February 1999.