ericaddison/dist_hw_2

UT Austin Distributed Systems HW 2

Distributed Systems HW 2

<UT Austin, Fall 2017> <Audrey Addison, Eric Addison>

Description

The programming portion of this HW was to implement a fault-tolerant distributed ticket-booking system. The list of "seat assignments" in this system acts as a shared resource, so read or write modifications to the list require access to a critical section of code. Access to the critical section is managed by Lamport's mutex algorithm.

Directory Structure

The source code does not use packages for modularity. The directory structure is:

• inputs: input config files for client and server
• lib: library directory containing Jackson JSON library for Java
• src: all source code for this HW

Key Code

Most of the code handles processing messages or TCP connections. Here are some key notes about the code:

• The core of the Lamport Mutex algorithm is implemented in a set of methods near the top of the LamportMutex class.
• Fault tolerance is achieved by:
  • Having the client attempt to reconnect to a different server if the connection is lost
  • Updating the number of live servers in the Mutex algorithm
  • Allowing servers to reconnect after going down

We did not address the issue of unresponsive servers ... only dead connections.

How to build

This code has one library dependency, which is a Java library for JSON serialization. This library must be included in the classpath. A script is included to compile the java code into a JAR file in build.sh, which runs the following commands:

mkdir build
javac -classpath ./lib/jackson-annotations-2.9.1.jar:./lib/jackson-databind-2.9.1.jar:./lib/jackson-core-2.9.1.jar:. -d ./build/ src/*.java

cd build
jar cvf ../HW2.jar *

cd ..
rm -rf build

How to run

Scripts are provided to run the Server or Client class from the compiled JAR file. They are called by:

./run_client.sh inputs/client1.txt          # run Client.main() with the given config file

./run_server.sh inputs/server1.txt          # run Server.main() with the give config file

./run_server.sh inputs/server2.txt restart  # restart a failed server (server 2)

The Servers would like to be started in order, however this is not necessary. All servers must start initially before a client can reliably connect. After the initial startup, servers can fail and come back up (using the restart option when launching), as long as one server is still running.