Why is multiaxo called multiaxo?
It is about several rounds of algorithm to reconcile sequential requests from a stable leader with minimal messages. In the beginning, without a recognized leader, you must run at least one round of basic Paxos where the candidate leader sends a request to prepare (using the terminology of the article Paxos Made Simple). Positive reviews from the majority confirm this as a leader. It then sends receive messages for that round, which succeeds if you receive most of the acknowledgments. Instead of starting with preparation requests, it can go straight into galloping mode, where it sends successive receive messages when it hears most of the acknowledgments for the previous receive request. This is very efficient as it requires a minimum number of messages, but it only occurs for a fewrounds from a stable leader. This can be interrupted by the leader knocking down yet another network glitch that causes the follower to time out on a healthy leader. He then issues his own training request as a management task, which is resolved by the Paxos core rules. Once you get a stable leader, he can go into galloping mode in several Paxos.
source to share
There are two phases in Paxo that require a lot of messages to exchange. To optimize the paxos protocol, we try to run the phase only when needed. The first phase talks about "election of a leader". It doesn't make sense to run this step over and over to get every log entry (state machine replication issue). We only run it if the leader fails. This is the time when the presenter is selected for the "log" and not just for each log entry. Multi-paxos means the same leader handles multiple client requests without having to elect a leader over and over again.
source to share