A node exhibits arbitrary behavior: it may transmit arbitrary messages at arbitrary times, take incorrect steps, or stop.

Byzantine failures occur when a node does not behave according to its specific protocol or algorithm. This usually happens when a malicious actor or a software bug compromises the node.

To cope with these failures, we need complex solutions. However, most companies deploy distributed systems in environments that they assume to be private and secure.

Fail-stop failures are the simplest and the most convenient ones from the perspective of someone that builds distributed systems. However, they are not very realistic. This is because there are many cases in real-life systems where it’s not easy for us to identify whether another node crashes or not.

Most of the algorithms we analyze in this course work under the assumption of crash failures.