1. 3 server machines
2. 1 client machine

1. Make sure Rabia is properly installed. Follow the instructions in the repo. This step is critical as it provides the go binary and python3.8 needed for testing.
2. SSH into each of the VMs and do the following:
   1. cd ~/go/src
   2. git clone https://github.com/zhouaea/epaxos.git && cd epaxos
   3. git checkout paxos-no-batching
   4. . compile.sh
   5. After compiling, you should see
      Built Master
Built Server
Built Client

For your convenience, we have created branches preloaded with the configurations we used for each experiment mentioned in the paper.

In each machine, git checkout into your desired branch. Compile go binaries on each machine with . compile.sh. Then, follow the instructions below, starting from Configure Machines.

git checkout paxos-no-batching

git checkout epaxos-no-batching

git checkout paxos-batching

git checkout epaxos-batching

git checkout paxos-batching-data-size-256B

git checkout epaxos-batching-data-size-256B

1. First, find the description of your machine by looking for a <node> tag with the attribute client_id=<your_machine_name>.
2. Inside the node tag, look for <interface>, and inside it you should see an <ip/> tag. The attribute address will give you the machine's experiment network ip.
3. The address likely has the format 10.10.1.x.

Inside base-profile.sh in each machine, configure the experimental network IP address of all server machines in the ServerIps array, client machines in the ClientIps array, and the server machine that will be the leader in the MasterIP variable.

1. Finally, run . runPaxos.sh or . runEPaxos.sh (depending on the branch) on your master machine only. You should see a relatively constant flow of messages in your terminal.
   1. Some of our results, namely those done for Figure 4 or Varying Data Size, require multiple experiments to be run with a varying number of clients. For these experiments, we provide 9 profiles, where profile0.sh runs 20 clients, while profile8.sh runs 500 clients. To change which profile is being executed, change the second line of either runPaxos.sh or runEPaxos.sh on each machine to execute your desired profile. By default, profile0.sh will be executed.
2. If all works correctly, there will be n client logs inside the /logs directory in your master machine.
3. For throughput/latency analysis, run:
   1. python3.8 analysis.py ./logs
   2. If you get the error below, you didn't supply ./logs as an argument

      Traceback (most recent call last):
           File "analysis.py", line 164, in <module>
             infos, disconnects = analysis_epaxos_logs()
           File "analysis.py", line 100, in analysis_epaxos_logs
             params = get_experiments(argv[1])
      IndexError: list index out of range

   3. If successful, this will output a variety of different statistics. You can find the relevant statistics as clientp50Latency: x, clientp99Latency: x, throughput: x .