A simple java implementation of the Gossip protocol. Messages between nodes are sent using protobuf. Nodes are multithreaded, and are capable of handling requests parallely.
The MyServer script is used to actually initialize a server. The usage is: ./MyServer <nodes_in_network> [ -i <input_file> ]
IT IS RECOMMENDED TO NOT RUN THIS SCRIPT MANUALLY
The Wrapper script initializes and runs all the servers based on what we give it. The usage is: ./Wrapper -p <number_of_type_1_processes> -n <number_of_type_2_processes> -i <input_file>
- Type 1 process is one that takes an input file and uses that to send messages (each line is considered a message), at regular intervals.
- Type 2 process is the process that has no input and just acts as a relay in the network.
Note: All Type 1 processes also have the functionality of Type 2 processes.
Each node has 2 default wait times defined:
- GOSSIP-READ-INPUT-TIME: The time that the process is supposed to wait between reading two lines of input.
- GOSSIP-PROCESS-TIME: The time taken to process a gossip. This is the delay done after calling the process gossip funtion.
The clock used in this simulation is very simple. The idea for the clock was to satisfy the following. If a process gets messages out of order, they should be discarded, two messages m1 and m2 are out of order if and only the below is true:
- m1 and m2 are generated by the same process, and
- m1 is read before m2 in the file, but the process has received m2 before m1
- This means the if m1 and m2 are generated by different process, they can never be out of order.
So the clock used is a combination of ProcessID:ProcessClock Look at the acceptMessage function to see how our clock comparision is done.