# Summary of Bipartisan Paxos — Part (2/6).

*This is part 2 of my summary of BPaxos. If you haven’t read part 1, you can find it **here**.*

In this post, I will focus on a brief summary of Fast Paxos. The rest of BPaxos protocols leverage Fast Paxos. Before I begin, I would say that I have not read the Fast Paxos paper, but I will focus on what the paper mentioned.

Fast Paxos is a two-phase consensus algorithm with a set of clients, leaders, acceptors, and learners. Fast Paxos proceeds in a series of integer-valued rounds with 0 being the smallest round and −1 being a null round. Every round is classified either as a fast round or as a classic round.

In phase 2a of Fast Paxos, a leader has to choose a value and to send it to the acceptors, because every intersection between fast quorum and quorum votes for the value proposed by the leader in the round k, **see Algorithm 1**.

The **key invariant** of Fast Paxos is that if a value v was maybe chosen in a round less than (I), then a leader must propose v in round (I). A leader can propose an arbitrary value in round I only if it has concluded that no value was chosen in any round less than i.

If we assume that round 0 is a fast round and every other round is a classic round, we can simplify the standard phase 2a algorithm shown in **Algorithm 1 **to the variant shown in **Algorithm 2**.

# End of Part 2

NEXT → Summary of Bipartisan Paxos (Part 3/6)