Key-Value Store Consistency

Replication is a key to providing high availability and fault tolerance in distributed systems. Here we implemented a Key-Value Store with replicated servers to meet these two important properties. Our implementation meets the correctness criteria of linearizability and sequential consistency.

Processes

1. server.py
2. client.py

Consistency Models

1. eventualConsistency.py
2. linearizabilityConsistency.py

To Run

Server:

usage: server.py [-h] id consistency W R

positional arguments:

        id           process id (1-10), default=1
        consistency  consistency model (eventual/linearizability), default=eventual
        W            number of w_ack indicates finished, default=1
        R            number of r_ack indicates finished, default=1

serverID = 1 indicates sequencer

Client:

usage: client.py [-h] serverID

positional arguments:

        serverID    server id, default=1

Note that log file 'output_log*.txt' will be created in the same directory.
New log info will be appended to the previous log file, you can delete
log files before start the new run.

Quick Start

1. Eventual Consistency:

Start servers at different terminals:
>> python server.py 1 eventual 2 2
>> python server.py 2 eventual 2 2

Start clients at different terminals:
>> python client.py 2
>> python client.py 2

2. Linearizability Consistency:

Start servers at different terminals:
>> python server.py 1 linearizability 2 2
>> python server.py 2 linearizability 2 2

Start clients at different terminals:
>> python client.py 2
>> python client.py 2

3. Test server crash:

Start servers at different terminals:
>> python server.py 1 eventual 2 2
>> python server.py 2 eventual 2 2

Start clients at different terminals:
>> python client.py 2

Then kill server2, after which client should be able to connect to
server1 automatically.

4. Test 'delay [ms]' to allow multipule command lines:
        Start servers at different terminals:
>> python server.py 1 eventual 2 2
>> python server.py 2 eventual 2 2

Start clients at different terminals:
>> python client.py 2
>> delay 10000
>> get x
>> put x 1
>> get x

5. Test 'dump' to show all variables status on server side:
>> python server.py 1 eventual 2 2
>> python server.py 2 eventual 2 2

Start clients at different terminals:
>> python client.py 2
>> dump