This project implements a distributed dataflow analysis program that can be run on multiple machines and processes. This program allows us to analyze large programs that may not fit on a single machine by distributing the work across multiple machines that communicate with eachother and a primary machine.

Our implementation is written in Scala and is designed to analyze JVM bytecode.

• MacOS
• Scala3
  • To install/upgrade using brew

  brew install coursier/formulas/coursier && cs setup

Simply navigate to the root folder and run

sbt "run path/to/java/dir"

We have included 3 test java programs in the src/test/testFiles directory. An example of running the program on a small example is

sbt "run src/test/testFiles/espTest"

To run this analysis on your own Java files, make sure to compile your Java code and just input the path to your directory.

We have a comprehensive suite of unit tests for individual parts of the codebase. These can be found in src/test/scala and can be run with

sbt test

├── analysis
├── project
├── src
│   ├── main
│   │   ├── protobuf
│   │   ├── scala
│   │   │   ├── analyses
│   │   │   ├── cfg
│   │   │   ├── cli
│   │   │   ├── dataflow
│   │   │   ├── lattice
│   ├── test
│   │   ├── scala
│   │   ├── testFiles
├── target
└── build.sbt

• The analysis folder contains the outputs of tcpdump which keeps track of all communication between machines in our program. These outputs are in binary form in the .pcap files and converted to ASCII in the .txt files. Python analysis of these files for message size can be found in this folder.
• Both the project and target folders are generated by sbt which is the interactive build tool for Scala and Java projects. These should not be manually modified.
• The directory src contains the main source code.
  • main/protobuf contains all the proto definitions of the servers and messages that we use.
  • main/scala contains the source scala code.
    • analyses contains the code for performing constant propogation analysis.
    • cfg contains all code related to the constructing the control flow graph from java bytecode and performing operations on the graph.
    • cli contains code for handling the CLI arguments.
    • dataflow contains the code specifying the worker machine implementations, and handles how machines communicate computation units to eachother.
    • lattice contains helper class definitions and utils.
  • test/scala contains unit tests.
  • test/testFiles contains 3 example Java programs that don't do anything meaningful but demonstrate how our program works across programs of different sizes and complexities.
• build.sbt is the config for our project and specifies the dependencies.

Our design notebook can be found here.