Definitions of each abbreviation in the diagram are;
m: Request message objectc: Client IDt: Timestampv: View IDn: Sequence IDi: Peer(Node) IDr: Result of the request's operation
In the diagram, the peer change its state to prepared or committed when the count value, which is the number of verified messages from other peers, is larger than 2.
Actually, the condition is count >= 2*f where f is the maximum number of faulty peers, which the network can tolerate. In this case, f is just 1, so the condition is count >= 2.
Every node replies the result of the request's operation to the client individually. The client will collect these reply messages and if f + 1 valid reply messages are arrived, the client will accept the result.
In this sample implementation, there is no client. So, every node including the primary will return its reply message to the primary.
Apache 2.0