Augmented Traffic Control: A tool to simulate network conditions

Overview

Augmented Traffic Control

build-status-image pypi-version

Full documentation for the project is available at http://facebook.github.io/augmented-traffic-control/.

Overview

Augmented Traffic Control (ATC) is a tool to simulate network conditions. It allows controlling the connection that a device has to the internet. Developers can use ATC to test their application across varying network conditions, easily emulating high speed, mobile, and even severely impaired networks. Aspects of the connection that can be controlled include:

  • bandwidth
  • latency
  • packet loss
  • corrupted packets
  • packets ordering

In order to be able to shape the network traffic, ATC must be running on a device that routes the traffic and sees the real IP address of the device, like your network gateway for instance. This also allows any devices that route through ATC to be able to shape their traffic. Traffic can be shaped/unshaped using a web interface allowing any devices with a web browser to use ATC without the need for a client application.

ATC is made of multiple components that interact together:

  • atcd: The ATC daemon which is responsible for setting/unsetting traffic shaping. atcd exposes a Thrift interface to interact with it.
  • django-atc-api: A Django app based on Django Rest Framework that provides a RESTful interface to atcd.
  • django-atc-demo-ui: A Django app that provides a simple Web UI to use atc from a mobile phone.
  • django-atc-profile-storage: A Django app that can be used to save shaping profiles, making it easier to re-use them later without manually re-entering those settings.

By splitting ATC in sub-components, it make it easier to hack on it or build on top of it. While django-atc-demo-ui is shipped as part of ATC's main repository to allow people to be able to use ATC out of the box, by providing a REST API to atcd, it makes it relatively easy to interact with atcd via the command line and opens the path for the community to be able to build creative command line tools, web UI or mobile apps that interact with ATC.

ATC architecture

Requirements

Most requirements are handled automatically by pip, the packaging system used by ATC, and each ATC package may have different requirements and the README.md files of the respective packages should be checked for more details. Anyhow, some requirements apply to the overall codebase:

  • Python 2.7: Currently, ATC is only supported on python version 2.7.
  • Django 1.10: Currently, ATC is only supported using django version 1.10.

Installing ATC

The fact that ATC is splitted in multiple packages allows for multiple deployment scenarii. However, deploying all the packages on the same host is the simplest and most likely fitting most use cases.

To get more details on how to install/configure each packages, please refer to the packages' respective READMEs.

Packages

The easiest way to install ATC is by using pip.

pip install atc_thrift atcd django-atc-api django-atc-demo-ui django-atc-profile-storage

Django

Now that we have all the packages installed, we need to create a new Django project in which we will use our Django app.

django-admin startproject atcui
cd atcui

Now that we have our django project, we need to configure it to use our apps and we need to tell it how to route to our apps.

Open atcui/settings.py and enable the ATC apps by adding to INSTALLED_APPS:

INSTALLED_APPS = (
    ...
    # Django ATC API
    'rest_framework',
    'atc_api',
    # Django ATC Demo UI
    'bootstrap_themes',
    'django_static_jquery',
    'atc_demo_ui',
    # Django ATC Profile Storage
    'atc_profile_storage',
)

Now, open atcui/urls.py and enable routing to the ATC apps by adding the routes to urlpatterns:

...
...
from django.views.generic.base import RedirectView
from django.conf.urls import include

urlpatterns = [
    ...
    # Django ATC API
    url(r'^api/v1/', include('atc_api.urls')),
    # Django ATC Demo UI
    url(r'^atc_demo_ui/', include('atc_demo_ui.urls')),
    # Django ATC profile storage
    url(r'^api/v1/profiles/', include('atc_profile_storage.urls')),
    url(r'^$', RedirectView.as_view(url='/atc_demo_ui/', permanent=False)),
]

Finally, let's update the Django DB:

python manage.py migrate

Running ATC

All require packages should now be installed and configured. We now need to run the daemon and the UI interface. While we will run ATC straight from the command line in this example, you can refer to example sysvinit and upstart scripts.

atcd

atcd modifies network related settings and as such needs to run in privileged mode:

sudo atcd

Supposing eth0 is your interface to connect to the internet and eth1, your interface to connect to your lan, this should just work. If your setting is slightly different, use the command line arguments --atcd-wan and --atcd-lan to adapt to your configuration.

ATC UI

The UI on the other hand is a standard Django Web app and can be run as a normal user. Make sure you are in the directory that was created when you ran django-admin startproject atcui and run:

python manage.py runserver 0.0.0.0:8000

You should now be able to access the web UI at http://localhost:8000

ATC Code Structure

ATC source code is available under the atc directory, it is currently composed of:

  • atc_thrift the thrift interface's library
  • atcd the ATC daemon that runs on the router doing the traffic shaping
  • django-atc-api A django app that provides a RESTful interface to atcd
  • django-atc-demo-ui A django app that provides a simple demo UI leveraging the RESTful API
  • django-atc-profile-storage A django app that allows saving shaping profiles to DB allowing users to select their favorite profile from a list instead of re-entering all the profile details every time.

The chef directory contains 2 chef cookbooks:

  • atc A cookbook to deploy ATC. It also allows to deploy ATC in a Virtual Box VM in order to develop on ATC.
  • atclient Set up a Linux Desktop VM that can be used to test shaping end to end.

atcd

atcd is the daemon that runs on the router that does the shaping. Interaction with the daemon is done using thrift. The interface definition can be found in atc_thrift.thrift.

atc_thrift

atc_thrift defines the thrift interface to communicate with the atcd daemon.

django-atc-api

django-atc-api is a django app that provide a REST API to the atcd daemon. Web applications, command line tools can use the API in order to shape/unshape traffic.

django-atc-demo-ui

django-atc-demo-ui is a simple Web UI to enable/disable traffic shaping. The UI is mostly written in React

django-atc-profile-storage

django-atc-profile-storage allows saving profiles to DB. A typical use case will be to save a list of predefined/often used shaping settings that you want to be able to accessing in just a few clicks/taps.

Developing on ATC

To make ATC development easier, we use Virtual Box and Vagrant to provision and run a VM that will run the ATC daemon and the ATC UI from your git checkout.

Interacting with ATC will only shape the traffic within the VM and not on the host.

Setting up the environment

Note: vagrant is an easy way to set up a test environment, but virtualization will produce different results than a setup on bare-metal. We recommend using vagrant only for testing/development and using bare-metal for setups which require realistic shaping settings.

You will need to install VirtualBox, Vagrant and a couple of plugins:

  • VirtualBox
  • Vagrant
  • Chef DK
  • Install some vagrant plugins:
  • vagrant plugin install vagrant-berkshelf --plugin-version '>= 2.0.1'
  • vagrant plugin install vagrant-omnibus
  • Clone this repo: git clone [email protected]:facebook/augmented-traffic-control.git atc

Running ATC

Once in the repo, go to the chef/atc directory and run:

vagrant up trusty

This will take some time before it completes, once the VM is provision, SSH into it:

vagrant ssh trusty

You should now be able to access ATC at: http://localhost:8080/

Using the Sample Profiles

Once you've got ATC up and running, you can run the script utils/restore-profiles.sh to setup the set of default profiles.

The script needs to be passed a hostname:port with the location of your ATC instance:

utils/restore-profiles.sh localhost:8080

After doing this, you should see the 10 sample profiles listed below in your ATC instance:

  • 2G - Developing Rural
  • 2G - Developing Urban
  • 3G - Average
  • 3G - Good
  • Cable
  • DSL
  • Edge - Average
  • Edge - Good
  • Edge - Lossy
  • No Connectivity

Naturally, you cannot improve your natural network speed by selecting a faster profile than your service. For example, selecting the Cable profile will not make your network faster if your natural connection speed resembles DSL more closely.

Hacking on the code

Hacking on ATC is done from the host and tested in the VM. In order to reflect the changes, you will need to start the services manually.

Both atcd and atcui have their python libraries installed in a python virtualenv so you will need to activate the environment in order to be able to run the services.

The virtualenv is installed in /usr/local/atc/venv/bin/activate .

source /usr/local/atc/venv/bin/activate

Running the daemon

The atcd daemon is running under the root user privileges, all operations below needs to be done as root.

To run the daemon manually, first make sure it is not running in the background:

service atcd stop

And run the daemon:

atcd

Once you are happy with your changes and you want to test them, you will need to kill the daemon and restart it in order to apply the changes.

Running the API/UI

This is a django project and, when running the django built-in HTTP server, will detect code changes and reload automatically.

To run the HTTP REST API and UI:

cd /var/django && python manage.py runserver 0.0.0.0:8000
Owner
Meta Archive
These projects have been archived and are generally unsupported, but are still available to view and use
Meta Archive
Official Pytorch Implementation of: "Semantic Diversity Learning for Zero-Shot Multi-label Classification"(2021) paper

Semantic Diversity Learning for Zero-Shot Multi-label Classification Paper Official PyTorch Implementation Avi Ben-Cohen, Nadav Zamir, Emanuel Ben Bar

28 Aug 29, 2022
Detectron2-FC a fast construction platform of neural network algorithm based on detectron2

What is Detectron2-FC Detectron2-FC a fast construction platform of neural network algorithm based on detectron2. We have been working hard in two dir

董晋宗 9 Jun 06, 2022
Implements Gradient Centralization and allows it to use as a Python package in TensorFlow

Gradient Centralization TensorFlow This Python package implements Gradient Centralization in TensorFlow, a simple and effective optimization technique

Rishit Dagli 101 Nov 01, 2022
Preprocessed Datasets for our Multimodal NER paper

Unified Multimodal Transformer (UMT) for Multimodal Named Entity Recognition (MNER) Two MNER Datasets and Codes for our ACL'2020 paper: Improving Mult

76 Dec 21, 2022
Exploring the Dual-task Correlation for Pose Guided Person Image Generation

Dual-task Pose Transformer Network The source code for our paper "Exploring Dual-task Correlation for Pose Guided Person Image Generation“ (CVPR2022)

63 Dec 15, 2022
[arXiv] What-If Motion Prediction for Autonomous Driving ❓🚗💨

WIMP - What If Motion Predictor Reference PyTorch Implementation for What If Motion Prediction [PDF] [Dynamic Visualizations] Setup Requirements The W

William Qi 96 Dec 29, 2022
Deep learning models for change detection of remote sensing images

Change Detection Models (Remote Sensing) Python library with Neural Networks for Change Detection based on PyTorch. ⚡ ⚡ ⚡ I am trying to build this pr

Kaiyu Li 176 Dec 24, 2022
Deep Ensemble Learning with Jet-Like architecture

Ransomware analysis using DEL with jet-like architecture comprising two CNN wings, a sparse AE tail, a non-linear PCA to produce a diverse feature space, and an MLP nose

Ahsen Nazir 2 Feb 06, 2022
AlgoVision - A Framework for Differentiable Algorithms and Algorithmic Supervision

NeurIPS 2021 Paper "Learning with Algorithmic Supervision via Continuous Relaxations"

Felix Petersen 76 Jan 01, 2023
HistoSeg : Quick attention with multi-loss function for multi-structure segmentation in digital histology images

HistoSeg : Quick attention with multi-loss function for multi-structure segmentation in digital histology images Histological Image Segmentation This

Saad Wazir 11 Dec 16, 2022
People movement type classifier with YOLOv4 detection and SORT tracking.

Movement classification The goal of this project would be movement classification of people, in other words, walking (normal and fast) and running. Yo

4 Sep 21, 2021
Reviving Iterative Training with Mask Guidance for Interactive Segmentation

This repository provides the source code for training and testing state-of-the-art click-based interactive segmentation models with the official PyTorch implementation

Visual Understanding Lab @ Samsung AI Center Moscow 406 Jan 01, 2023
PyTorch implementation for COMPLETER: Incomplete Multi-view Clustering via Contrastive Prediction (CVPR 2021)

Completer: Incomplete Multi-view Clustering via Contrastive Prediction This repo contains the code and data of the following paper accepted by CVPR 20

XLearning Group 72 Dec 07, 2022
A Shading-Guided Generative Implicit Model for Shape-Accurate 3D-Aware Image Synthesis

A Shading-Guided Generative Implicit Model for Shape-Accurate 3D-Aware Image Synthesis Figure: Shape-Accurate 3D-Aware Image Synthesis. A Shading-Guid

Xingang Pan 115 Dec 18, 2022
Offline Multi-Agent Reinforcement Learning Implementations: Solving Overcooked Game with Data-Driven Method

Overcooked-AI We suppose to apply traditional offline reinforcement learning technique to multi-agent algorithm. In this repository, we implemented be

Baek In-Chang 14 Sep 16, 2022
Pytorch implementation of the unsupervised object discovery method LOST.

LOST Pytorch implementation of the unsupervised object discovery method LOST. More details can be found in the paper: Localizing Objects with Self-Sup

Valeo.ai 189 Dec 25, 2022
Ranger - a synergistic optimizer using RAdam (Rectified Adam), Gradient Centralization and LookAhead in one codebase

Ranger-Deep-Learning-Optimizer Ranger - a synergistic optimizer combining RAdam (Rectified Adam) and LookAhead, and now GC (gradient centralization) i

Less Wright 1.1k Dec 21, 2022
Pytorch implementation of our paper under review — Lottery Jackpots Exist in Pre-trained Models

Lottery Jackpots Exist in Pre-trained Models (Paper Link) Requirements Python = 3.7.4 Pytorch = 1.6.1 Torchvision = 0.4.1 Reproduce the Experiment

Yuxin Zhang 27 Jun 28, 2022
Automatic meme generation model using Tensorflow Keras.

Memefly You can find the project at MemeflyAI. Contributors Nick Buukhalter Harsh Desai Han Lee Project Overview Trello Board Product Canvas Automatic

BloomTech Labs 2 Jan 13, 2022
Official Pytorch Implementation of: "ImageNet-21K Pretraining for the Masses"(2021) paper

ImageNet-21K Pretraining for the Masses Paper | Pretrained models Official PyTorch Implementation Tal Ridnik, Emanuel Ben-Baruch, Asaf Noy, Lihi Zelni

574 Jan 02, 2023