Skip to content

Latest commit

 

History

History
101 lines (75 loc) · 3.98 KB

README.md

File metadata and controls

101 lines (75 loc) · 3.98 KB
Raw

LinkGuardian Simulator

This repository includes the simulator used in the evaluation of our SIGCOMM'23 paper - LinkGuardian (section 4.8). This open-source simulator is designed based on the closed-source simulator used by Zhuo et al. and described in the SIGCOMM'17 paper - CorrOpt (section 7.1).

If you haven't read our paper, we provide a brief summary of this simulator's design here.

Installation and Usage

Note for Artifact Evaluation

If you want to simply run the simulator for artifact evaluation, we provide a convenient Docker container image. Please refer to the README and Makefile in the sigcomm23-ae/ directory.

If you want to have a local installation of the simulator, for example, to tweak it or build on top of it, please continue reading.

Local Installation

Note All the steps below assume that you are in the root directory of this repository that contains the main.py file. These steps have been tested and are expected to work on Ubuntu 18.04 and above.

The simulator is implemented in Python3 and requires the following packages:

pylint networkx ipython jupyter termcolor matplotlib tabulate pandas tqdm progress

Instead of manually installing these packages, we recommend using a Python virtual environment managed by conda. If you do not already have conda, we recommend installing it through Miniconda which provides a minimal conda installation compared to Anaconda.

Once you have conda installed and available in your $PATH, you can create the required Python virtual environment (called lg-sim) using the following command (one-time step):

conda env create -f ./conda_env-lg_sim.yml

Local Usage

After the conda environment lg-sim is created, you can activate the same before running the simulator.

conda activate lg-sim

Once the lg-sim environment is activated, it should show you a prefix to your shell prompt as following:

(lg-sim) user@server:~/linkguardian-simulator$

Now, the simulator can be run as following by providing a config JSON file which consists of the required inputs for the simulator run:

./main.py ./simulation_configs/sigcomm23_eval/fbfabric_100k_os1-corropt_50.json

During a simulation run, the simulator first builds the topology (in-memory) and then plays the failure trace while applying the specified solution.

Output Files

Each simulator run produces two files in the output_dir as specified in the simulation config file:

  1. A data file (.dat) consisting of a timeseries of topology-level performance parameters; and
  2. A log file (.log) showing the logs from the simulation run.

Provided Simulation Configs and Failure Trace

We provide four different simulation configs used in our paper's evaluation under simulation_configs/sigcomm23_eval. The four configs result from the fact that we have two solutions (CorrOpt and LinkGuardian + CorrOpt) each to be run with two capacity constraints (50% and 75%).

All four configurations use a large FB Fabric topology consisting of ~100k switch-to-switch links and a corresponding 1-year long link failure trace provided here.

Trace Generation

The trace that was used in our paper's evaluation was generated using the trace generator provided in this repository and described in our paper (Appendix D). Please refer to trace_generator/ for more details.