Checkpointing in a Distributed Coordination Protocol for Multiple Peer Processes

There are many discussions on agreement protocols of multiple processes. However, they just aim at agreeing one value out of values shown by the processes. In human societies, agreement procedures are flexible, person can change their opinions and can use various type of agreement conditions. In this paper, we discuss a more flexible agreement protocol of multiple peer processes by taking account the human behaviour in social agreement procedures. Each process first takes a value v in a domain Di which is a collection of possible values and notifies the other processes of the value v. A process in turn receives values from other processes. Unless a set of the values from all the processes satisfy an agreement condition, a process can take another value v´. In order to model the social behaviour of people, we already discuss E- and P-precedent relations on values. In order to efficiently make an agreement, we have to discuss some coordination mechanism of multiple processes. A process pi would like to start over from a previous round if pi finds it to have selected an improper value. The process pi takes a checkpoint at a round. Then, the process pi can roll back to the previous checkpointed round. In addition, there are some constraints on a value. If a process says some value, the process cannot withdraw the value, i.e. cannot roll back. Based on the precedent relations and checkpoints, a group of processes efficiently make an agreement.