Violet uses selective symbolic execution to automatically model and detect bad performance configurations in large systems.
Violet: Selective Symbolic Execution to Detect Bad Performance Misconfiguration
Violet is designed for system administrators, developers, and security engineers who need to identify and assess specious or misconfigured performance settings in complex software systems. It automates the reasoning about configuration parameters to detect risky or harmful configurations that could degrade system performance or reliability.
Violet is tested primarily on Ubuntu 16.04; running on newer versions like 18.04 may require additional effort. The build process is resource-intensive, requiring over 10GB of disk space and considerable time. It is recommended to run Violet on physical machines rather than virtualized environments to avoid known issues. Users should consult the OSDI '20 paper for detailed technical design and cite it when using Violet in research.
Clone the repository: git clone https://github.com/OrderLab/violet.git
Initialize submodules: git submodule update --init --recursive
Add S2E local build path to PATH: export PATH=$HOME/.local/bin:$PATH
Re-login shell to apply PATH changes
Run build script: ./build.sh (tested on Ubuntu 16.04)
Build QEMU guest images as per instructions
Build target system following the documented stepsgit clone https://github.com/OrderLab/violet.git
Clones the Violet root repository to local machine.
git submodule update --init --recursive
Initializes and updates all submodules required by Violet.
export PATH=$HOME/.local/bin:$PATH
Adds the local S2E build directory to the system PATH.
./build.sh
Builds S2E and Violet components; a lengthy process requiring significant disk space.
Build QEMU guest images
Prepares virtual machine images for symbolic execution analysis.
Create an analysis project and initialize it
Sets up a new project for symbolic execution and configuration analysis.
Run symbolic execution
Executes the symbolic execution engine to derive performance models.
Build and run trace analyzer
Analyzes collected execution traces to detect specious configurations.
Run static analyzer
Extracts related configuration files and metadata from the target system.
Re-run symbolic execution with related configurations
Performs targeted symbolic execution based on static analysis results.