proxybench

Benchmark different proxies on basis of the following:

test	Implemented	Description
HTTP Stress	✅	records the time it takes to finish the test on http://httpvshttps.com
HTTPS Stress (HTTP/2)	✅	records the time it takes to finish the test on https://httpvshttps.com
Heavy Websites	✅	records the time it takes to open websites like https://techcrunch.com/
HTTPS Stress (HTTP/1.1)
WS Support
WSS Support
<TO ADD MORE>

Installation

• go get github.com/pulkitsharma07/proxybench

Running the benchmark

• Make sure you have chromedriver present in your PATH.
• Start the desired proxies on different ports.
• Make sure you add and trust the certificates provided by the respective proxies, so that they can be considered as a trusted CA by the browser.
• Refer proxybench.json.sample and create proxybench.json similarly.
• Launch proxybench, it will look for proxybench.json in the current directory and run according to that config.

Reporting

Generates the following:

+--------------------------------+----------------+------------------------+
|          PROXY CONFIG          |   BENCHMARK    | COMPLETED IN (SECONDS) |
+--------------------------------+----------------+------------------------+
| Proxy: [Direct]                | HTTP Stress    | [15.575000]            |
| Proxy: [Direct]                | HTTPS Stress   | [2.757000]             |
| Proxy: [Direct]                | Heavy Websites | [20.703525]            |
| Proxy:                         | HTTP Stress    | [17.920000]            |
| [browsermob(littleproxy)]      |                |                        |
| Proxy:                         | HTTPS Stress   | [17.053000]            |
| [browsermob(littleproxy)]      |                |                        |
| Proxy:                         | Heavy Websites | [40.902084]            |
| [browsermob(littleproxy)]      |                |                        |
| Proxy: [browsermob(legacy)]    | HTTP Stress    | [20.777000]            |
| Proxy: [browsermob(legacy)]    | HTTPS Stress   | [17.399000]            |
| Proxy: [browsermob(legacy)]    | Heavy Websites | [31.580184]            |
| Proxy: [mitmproxy]             | HTTP Stress    | [39.376000]            |
| Proxy: [mitmproxy]             | HTTPS Stress   | [47.631000]            |
| Proxy: [mitmproxy]             | Heavy Websites | [198.824246]           |
| Proxy: [mitmdump]              | HTTP Stress    | [17.960000]            |
| Proxy: [mitmdump]              | HTTPS Stress   | [8.147000]             |
| Proxy: [mitmdump]              | Heavy Websites | [41.605729]            |
| Proxy: [charles]               | HTTP Stress    | [18.535000]            |
| Proxy: [charles]               | HTTPS Stress   | [6.322000]             |
| Proxy: [charles]               | Heavy Websites | [52.927620]            |
+--------------------------------+----------------+------------------------+

• Generated using: browsermob v2.1.4, mitmproxy/mitmdump v3.0.3, charles 4.2.1
• Direct stands for the baseline performance, i.e. without using any proxy.

Development

• Use dep to install dependencies
• Add proxy information in proxybench.json
• benchmarks are defined in benchmarks/
• You can define new benchmarks similarly, add them to NewSimpleSuite to execute.
• go install
• proxybench to run (Assuming GOBIN is added to your PATH)

TODO

• Unit and integration tests.
• CLI option to skip the baseline (DIRECT) test.
• CLI options to launch tests in sync/async.
• Add wrapper around Run to measure the proxy's CPU/Memory usage (Need to take PID from user, or figure out from script?).Eventually render a timeseries graph.
• Randomize test order and most importantly do multiple test runs and report on standard deviation. (Thanks @mhills)
• Test reports, generate test reports confining to some format, each test case will have corresponding timeseries graphs of CPU, memory, disk, number of sockets etc of the proxy.
• Final plan is to give every proxy a score based on its performance in comparison to the baseline for the different benchmarks.
• Launch proxies from this script?

PS: I made this mainly to learn Go. I will really appreciate if you can point out things in the code which I should improve upon. For example naming conventions, directory structure, or new benchmarks etc. Feel free to open any issues.

Thank You !