hmc-clinic-msft-2020

Contains all work from the Harvey Mudd College 2020-2021 Microsoft Clinic team.

This repository currently contains

1. Install scripts for installing a NGINX Server with a patch for Cloudflare's Quiche implementation of HTTP/3, a Caddy Server
2. Scripts for testing HTTP/3 and HTTP/2 with simulated network conditions.
3. Scripts for installing Web Page Test!.
4. Scripts for seting up a client VM and a server VM on Azure.
5. A Script for creating a network namespace to run experiments in
6. Dummy files as the payloads of the servers.
7. A Script for visualization the database.

This is designed to run on Ubuntu 18.04.

Install

To set up a client VM, run the following in the desired directory:

git clone https://github.com/JBoerma/hmc-clinic-msft-2020.git
cd hmc-clinic-msft-2020/install/
./install-clients.sh

Similarly, to set up a server VM:

git clone https://github.com/JBoerma/hmc-clinic-msft-2020.git
cd hmc-clinic-msft-2020/install/
./install-servers.sh 

Running the Script

This script will access a given URL using HTTP/2 and HTTP/3 under specified network conditions and with specific browsers, and return navigation timings using the PerformanceNavigationTiming interface, which implements the Performance Timing Level 2 specification. The script currently uses tc-netem to modify the network conditions, so root access is required.

With Private Server Virtual Machine

Ensure that

• you have added a file ips.json on the Client VM with the public and private IPs of the Server VM. Follow the example of ips.example.json
• you have a private key in a file MSFT_Clinic_Key.pem

This will enable the script to automatically start servers on the Server VM, as well as enable monitoring of server resources.

Dependencies

• Python 3.6+
• playwright-python

Usage

experiment.py [--device DEVICE] [--conditions CONDITIONS ...] [--browsers BROWSERS ...] [--url URL] [--runs RUNS] [--out OUT] [--throughput THROUGHPUT] [--payloads PAYLOADS] [--ports PORTS ...] [options]

Arguments

--device DEVICE           Network device to modify [default: lo]
--conditions CONDITIONS   List of network conditions [default: 4g-lte-good 3g-unts-good]
--browsers BROWSERS       List of browsers to test [default: chromium edge]
--throughput THROUGHPUT   Maximum number of request to send at a time [default: 1]
--url URL                 URL to access [default: https://localhost]
--runs RUNS               Number of runs in the experiment [default: 100]
--out OUT                 File to output data to [default: results/results.db]
--ports PORTS             List of ports to use (':443', ':444', ':445', ':446') [default: :443]
--payloads PAYLOADS       List of sizes of the requsting payload [default: 100kb 1kb 10kb]

Options

-h --help                 Show this screen 
--disable_caching         Disables caching
--warmup                  Warms up connection
--async                   Run experiment asynchronously
--qlog                    Turns on QLog logging

For example, to access a specific server 10 times through Firefox with a good 4g network, run:

python3 experiment.py --browsers firefox --conditions 4g-lte-good --out results/data.db --url localhost --runs 10

The output of this script will be stored in two files:

In the client VM, results/data.db contains the sqlite database, that stores the requests parameters, network condition, and navigation timings.

In the server VM

Using a Network Namespace

If starting tests through SSH on a VM, the experiments can throttle your connection if network conditions are simulated on the same device your SSH connection is going through. To prevent this, there is a script install/namespace.sh which initializes network namespace with a virtual ethernet device to simulate conditions on.

Usage:

./install/namespace.sh
What network device do you want to use?
 lo:
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:
 wlo1:
    link/ether b4:69:21:e8:1e:fd brd ff:ff:ff:ff:ff:
 enx000ec6aeac34:
    link/ether 00:0e:c6:ae:ac:34 brd ff:ff:ff:ff:ff:
 docker0:
    link/ether 02:42:20:76:74:0e brd ff:ff:ff:ff:ff:

lo
lo

Running the script will initially print out all currently connected devices. Type the name of the desired device to run experiments on, and the script will create a new network namespace that can connect to other networks through that device.

In order to run experiments inside the namespace, run:

sudo ip netns exec netns0 sudo -u $USER python3 experiment.py --device veth-netns0

Followed by any desired experiment parameters.